Radiation Characterization of a Hardened 0.22 m Anti-Fuse Field Programmable Gate Array R.J. Nejad...

43
Radiation Characterization of a Hardened 0.22 m Anti- Fuse Field Programmable Gate Array R.J. Nejad 1 , P.A. Rickey 2 , K. Konadu 2 , W.J. Stapor 2 , P.T. McDonald 2 , W. Heidergott 3 1 Welkin Associates LTD., Chantilly VA 20151 2 Stapor Research Inc., Chantilly VA 20151 3 General Dynamics C4 Systems, Scottsdale AZ 85251 May 10, 2006

Transcript of Radiation Characterization of a Hardened 0.22 m Anti-Fuse Field Programmable Gate Array R.J. Nejad...

Page 1: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

Radiation Characterization of a Hardened 0.22 m Anti-Fuse Field

Programmable Gate Array

R.J. Nejad1, P.A. Rickey2, K. Konadu2, W.J. Stapor2, P.T. McDonald2, W. Heidergott3

1Welkin Associates LTD., Chantilly VA 201512Stapor Research Inc., Chantilly VA 20151

3General Dynamics C4 Systems, Scottsdale AZ 85251

May 10, 2006

Page 2: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 2

Outline

• Introduction• Target device personalities• Measurements methodology• Heavy ions• Protons• Data reduction and analysis• Summary

Page 3: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 3

Goal

VERIFY SUITABILITY OF UMC ACTEL RTSX-SU SERIES FPGA

DEVICES FOR USE IN LEO

Page 4: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 4

Background

• Change to UMC foundry devices required accurate SEE rate assessment

• SEE rates for MEC foundry devices were known– No empirical reason to expect that UMC devices have

the same rates

• Biggest problem for any new device is Destructive SEE (DSEE)– Not empirically verified with applicable particle radiation

and suitable operating conditions

• Limited existing NASA/Actel data was encouraging

Page 5: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 5

Limited Actel UMC SEU Data

10-11

10-10

10-9

10-8

0 10 20 30 40 50 60 70

Only 4 points in eachData is limited

LET < 30 MeV/(mg/cm2)

may need furthermeasurements

Actel RTSX UMC DevicesEarly NASA Heavy Ion Data from BNLSRI Analysis 10/04

0.5 MHz

10 MHz

LET (MeV/(mg/cm2))

CR

OS

S S

EC

TIO

N (

cm2 /b

it)

Page 6: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 6

Issue With Existing Data

10-10

10-9

10-8

0 10 20 30 40 50 60 70

Low LET behavioris undetermined

10 MHz Data with 90% Prediction Interval

Actel RTSX UMC DevicesEarly Heavy Ion Data from BNLSRI Analysis 11/04

LET (MeV/(mg/cm2))

CR

OS

S S

EC

TIO

N (

cm2 /b

it)

Page 7: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 7

Vendor Data

10-7

10-5

10-3

10-1

10 20 50 100 200

LET (MeV/(mg/cm2))

CR

OS

S S

EC

TIO

N (

cm2 /d

evi

ce)

Page 8: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 8

Subsequent Vendor Data

10-7

10-5

10-3

10-1

10 20 50 100 200

LET (MeV/(mg/cm2))

CR

OS

S S

EC

TIO

N (

cm2 /d

evi

ce)

Page 9: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 9

Path

• Developed suitable RTSX 32 & 72 target device personalities

• Operated at 33, 50, and 100 MHz and as-used biasing• TAMU Heavy ion irradiations

– Verify No Destructive SEE • Up to LET 40 MeV/(mg/cm2)

• Minimum 1E7 particles/cm2

– Measured & analyzed Non-Destructive SEE cross sections & rates

• IUCF Proton irradiations– Verify No Destructive SEE

• Up to 197 MeV

– Measured & analyzed Non-Destructive SEE cross sections & rates– Accounted for accumulated proton dose (TID) during irradiations

Page 10: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 10

Target Device Personalities

• Built from Actel standard library• Designed to reflect as-used configuration by

providing combinatorial logic between flip flops– Follows application logic architecture– Allows separable cross sections from particle radiation

• Organized as separate ‘threads’– Basic logic block typically consists of 10 buffers, flip

flops, or logic functions– Thread made of multiple concatenated blocks– Thread size is easily configurable– Our threads were tailored to cover application

functionality

Page 11: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 11

Basic Test Circuits

• 72S test circuits are a superset of the 32S circuits• Common 72S & 32S threads

– 400 R-cell flip flops (DF1CB)– 400 C-cell flip flops (DF1CB_CC)– 400 R-cell flip flops (DF1CB) & 400 C-cell 2 level

combinatorial logic (CM8)– I/O Buffer thread

• 20 R-cell flip flops• 80 Input• 80 Output cells• external loop

Page 12: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 12

Common Target Blocks

DF1CB (A)

DF1CB_CC (B)

I/O (C)DF1CB & CM8 (F)

Combinatorial logic

Page 13: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 13

Target Distribution

SX72 and SX32 targets loaded up and running at speed

RT54SX32S RT54SX72SLogic

Sequential 886 of 1080 (82%) 1856 of 2012 (92%)Combinatorial 1608 of 1800 (90%) 2484 of 4024 (62%)

IO & Clocks 181 of 224 (80%) 199 of 209 (95%)HCLOCK 1 OF 1 1 OF 1

CLOCK 2 OF 2 2 OF 2QCLOCK -- 4 OF 4

Max Clock Freq.HCLOCK 205 MHz 193 MHz

CLOCK 146 MHz 139 MHzQCLOCK -- 111 MHz

DELAY 331 ns 207 ns

Page 14: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 14

Design Advantages

• Used long shift registers to increase sample size– Compared to previous measurements

• Error counting is automated (10 channels simultaneously)

• Separable SEE cross section data• SEE susceptibility measured at speed

– R-cell– C-cell flip flop

• Combinatorial SEE rate characterized vs frequency

• I/O-cell SEE characterized• Clock SEE

Page 15: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 15

SX72 Measurement Circuit Summary

REF LETTER CIRCUIT FUNCTION TEST EQUIPMENT

A R Cell Flip Flop Test CircuitDF1CB

Chain of 400 R Cell flip flops used to measure R Cell cross section

BERT

B C Cell Flip Flop Test CircuitDF1CB_CC

Chain of 400 C Cell flip flops BERT

C I/O Buffer Test Circuit Chain of 80 R Cell flip flops interleaved with 80 Input & 80 Output blocks with external pin looping

BERT

D C Cell Test Circuit DF1CB, XOR3 & BUF

Chain of R Cell flip flops interleaved with C Cell combinatorial logic

BERT

E C Cell TestDFC1B, XOR3B

Chain of R Cell flip flops interleaved with C Cell 1 level combinatorial logic

BERT

F C Cell Test DF1CB, CM8

Chain of 400 R Cell flip flops interleaved with 400 C Cell 2 level combinatorial logic

BERT

G R Cell TMR DF1CB

Chain of TMR R Cell Flip Flops with 3 R Cells and 1 C Cell majority voter

BERT

H R Cell EnableDFE3C

Chain of R-cell flip flops with enable BERT

J, K Counter Two counters and comparators Event Counter

L Clock Test Circuit HCLK Monitors clock networks for runt pulses Oscilloscope

M Clock Test Circuit CLKA Monitors clock networks for runt pulses Oscilloscope

N Clock Test Circuit QCLK Monitors clock networks for runt pulses Oscilloscope

P Buffer Test Circuit Chain of buffers used to measure changes in propagation delay with TID

Signal Generator

Oscilloscope, (offline)

Page 16: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 16

Supply Voltage & IO Configuration

• Test design is configured for 2.5V core voltage and 3.3V IO voltage

• All Inputs are configured as– LVTTL logic threshold– Power Up State ‘none’ (i.e. Neither a pull-up nor a pull-

down resistor)

• All Outputs are configured as– LVTTL standard– High slew– Power Up State ‘none’ (i.e. Neither a pull-up nor a pull-

down resistor)– 35 pF of capacitive loading

Page 17: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 17

SX72 Target Dimensions

1.5 cm × 0.8 cm die

Page 18: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 18

Measurement Methodology

• Measured bit errors in each thread using a PRBS input sequence– Developed a Xilinx FPGA based BERT to

simultaneously exercise all test threads– Scaled to 10 receiver channels

– FMAX > 100 MHz

– RS-232 Command & Control– Uses 27-1 or 210-1 PRBS sequence (210-1 used)

• Measured supply current and propagation delay for TID degradation– Power supply meter for current– Signal generator and oscilloscope for prop delay

Page 19: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 19

Measurement Setup

DUT

TARGET CAVE

RECEIVERS DRIVERS

USER CAVE

PWBPWB

Xilinx

BERT

&

EVENT

COUNT

2X 2X 10X

10X

O’SCOPELAPTOP

LAN

RS232

SIG GEN

DATA RATE

SIG GEN

COMM RATE

PC

4 4

10 10

Page 20: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 20

TAMU Beams

Ion Z AEnergy

(MeV/amu)

Kinetic Energy

(MeV) dE/dx in (Si) (MeV/(mg/cm2))

Range (Si) (μm)

129Xe 54 129 22.66 2923 0.217 39.6 255

84Kr 36 84 15.45 1298 0.180 25 177

84Kr 36 84 23.12 1942 0.219 20.1 302

40Ar 18 40 3.85 154 0.091 15 40

40Ar 18 40 9.60 384 0.142 10 123

40Ar 18 40 23.80 952 0.222 5.6 463

Page 21: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 21

Measurement Equipment

DUT

BERTBERT

INTERFACE

Page 22: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 22

TAMU Target Beam Alignment

Page 23: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 23

TAMU User Cave

Page 24: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 24

IUCF Measurements SetupBeam Alignment Equipment Cart

Page 25: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 25

Analysis

• For each test thread (heavy ions & protons)– Channel Cross section per bit

• (# of errors) / (particle fluence)

– Extracted Cross sections• R Cell, CC Cell, C Cell Combinatorial, IOBUF• Function of LET, Cell Type, and Clock Rate

• Propagation delay (thread P)– Fluence– Total Ionizing Dose (protons)

• DUT supply current– Fluence– Total Ionizing Dose (protons)

Page 26: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 26

Cross Section Extraction

RAW Channel

Measurement Data

Matrix

Extraction

Channel

RUN

# EventsRUN

FluenceRUN

Run R Cell

C Cell

IO Cell

CC Cell

Statistical

Combination

Mean CELL and

Uncertainty

Page 27: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 27

Thread Cell Distribution

Circuit Threads

BERT Channel

R Cell Count

C Cell Count

I/O Buffer

CountCC Flip Flop

Count

IOBUF 7 80 1* 163* 0

DFC1B 3 400 0 2 0

DFC1B_CM8 6 400 400 2 0

DFC1B_CC 4 0 0 2 400**

* IOBUF chain contains a MUX (C Cell) and 3 I/O Buffers in the circuit to enable/disable the chain

** DFC1B contains 800 C Cells configured as 400 CC Flip Flops

• Four circuit threads common to both 32SX and 72SX targets were used to extract R Cell, C Cell, IO Cell and CC Cell

Page 28: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 28

Matrix Math

• 4 unknowns– R Cell cross section, R– I/O Buffer cross section, I/O– C Cell cross section, C– CC Flip Flop cross section, CC

• 4 test threads– DFC1B (thread A)– DFC1B_CC (thread B)– DFC1B_CM8 (thread F)– IOBUF (thread C)

• Together they form 4 unknowns and 4 equations– Equations that describe the threads can be written in matrix form

• Matrix algebra solution

Page 29: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 29

Matrix Form

CC

C

I

R

400020

04002400

002400

0116380

DFC1B_CC

DFC1B_CM8

DFC1B

IOBUF

DFC1B_CC

DFC1B

IOBUF

400020

04002400

002400

0116380

CC

C

I

R

DFC1B_CM8

1

DFC1B_CC

DFC1B

IOBUF

32.5E87.69E66.07E53.08E

032.5E32.5E0

051.54E31.21E36.15E

087.69E32.51E53.08E

CC

C

I

R

DFC1B_CM8

Page 30: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 30

Cell Data Cross Sections

• Heavy ion cross sections measured for a range of LET values from ~5 to 40 MeV/(mg/cm2)

• No clear LET thresholds observed– No need to include LET threshold– No need for traditional and somewhat limited Weibul

• Cell cross section data analyzed using a power law function ( = A LETB)– Conservative approach– Function has the key features of the data and fits well

for IO, C, CC, and R Cell datasets– Accounts for potentially small but non-zero contributions

from very low LET values• Includes 0,0 data point

Page 31: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 31

Comments on Multiple Bit Errors

• Multiple bit errors (MBE) are defined as consecutive bits in error from a single particle strike

• Very few MBE were observed on all channels for clock rates < 100 MHz

• Most MBE observed at 100 MHz– Especially in the IOBUF channel– Interaction between SET width and clock width– System and design specific issue– All bits in error were treated as single-bit SEU for

conservative rate estimates in this analysis

Page 32: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 32

Sample MBE Distribution @ 33 MHzRun 14, Type 72, LET 24.8 MeV/(mg/cm2)

Summary Statistics================== (bits) (err/s) (s/chan) (bits) (bits) (bits) (bits)CH ERRORS SYNCH RESET ROLLOVER RATE(/s) ELAPSED 1-BIT 2-BIT 3-BIT 4>-BIT-- ------ ----- ----- -------- -------- ------- ------ ------ ------ ------ 1 2 1 0 0 0.25E-02 813 2 0 0 0 2 4 1 0 0 0.49E-02 813 4 0 0 0 3 4 1 0 0 0.49E-02 813 4 0 0 0 4 27 1 0 0 0.33E-01 813 27 0 0 0 5 0 1 0 0 0.00E+00 813 0 0 0 0 6 4 1 0 0 0.49E-02 813 4 0 0 0 7 0 1 0 0 0.00E+00 813 0 0 0 0 8 0 1 0 0 0.00E+00 813 0 0 0 0 9 1 1 0 0 0.12E-02 813 1 0 0 010 1 1 0 0 0.12E-02 813 1 0 0 0

Page 33: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 33

Sample MBE Distribution @ 100 MHzRun 12, Type 72, LET 24.8 MeV/(mg/cm2)

Summary Statistics================== (bits) (err/s) (s/chan) (bits) (bits) (bits) (bits)CH ERRORS SYNCH RESET ROLLOVER RATE ELAPSED 1-BIT 2-BIT 3-BIT 4>-BIT-- ------ ----- ----- -------- -------- ------- ------ ------ ------ ------ 1 24 1 0 0 0.28E-01 868 20 2 0 0 2 29 1 0 0 0.33E-01 868 29 0 0 0 3 46 1 0 0 0.53E-01 868 44 1 0 0 4 123 1 0 0 0.14E+00 868 113 5 0 0 5 4 1 0 0 0.46E-02 868 0 2 0 0 6 51 1 0 0 0.59E-01 868 51 0 0 0 7 333 1 0 0 0.38E+00 868 32 59 32 16 8 19 1 0 0 0.22E-01 868 19 0 0 0 9 11 1 0 0 0.13E-01 868 5 3 0 010 12 1 0 0 0.14E-01 868 6 3 0 0

Page 34: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 34

All R Cell Data

10-11

10-10

10-9

10-8

10-7

0 10 20 30 40 50 60 70 80 90 100

0.5 MHz NASA Data

10 MHz NASA Data

50 MHz

100 MHzUMC Actel RTSX32/72-SU FPGATAMU Heavy IonsR CELL Power Law Analysis

SRI 2/2005

LET (MeV/(mg/cm2))

CR

OS

S S

EC

TIO

N (

cm2 /C

EL

L)

Page 35: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 35

R CELL 10, 50, & 100 MHz

10-13

10-11

10-9

10-7

10-5

0 10 20 30 40 50 60

10 MHz

50 MHz

100 MHz

= ALETB

100 MHzA = 1.31E-13B = 3

50 MHzA = 3.43E-14B = 3

10 MHzA = 6.96E-15B = 3

UMC Actel RTSX32/72-SU FPGATAMU Heavy IonsR CELL Power Law

SRI 9/2005

LET (MeV/(mg/cm2))

CR

OS

S S

EC

TIO

N (

cm2 /C

EL

L)

Page 36: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 36

C CELL 10, 50, & 100 MHz

10-13

10-11

10-9

10-7

10-5

0 10 20 30 40 50 60

= ALETB

10 MHz

50 MHz

100 MHz

100 MHzA = 5.60E-14B = 3

50 MHzA = 1.60E-14B = 3

10 MHzA = 2.82E-15B = 3

UMC Actel RTSX32/72 FPGATAMU Heavy IonsC CELL Power Law Analysis

SRI 9/2005

LET (MeV/(mg/cm2))

CR

OS

S S

EC

TIO

N (

cm2 /C

EL

L)

Page 37: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 37

IO Buffer 10, 50, & 100 MHz

10-13

10-11

10-9

10-7

10-5

0 10 20 30 40 50 60

10 MHz

50 MHz

100 MHz

= ALETB

100 MHzA = 7.5E-12B = 3

50 MHzA = 5.1E-14B = 3

10 MHzA = 8E-15B = 3

UMC Actel RTSX32/72-SU FPGATAMU Heavy IonsIO Buffer Power Law

SRI 9/2005

LET (MeV/(mg/cm2))

CR

OS

S S

EC

TIO

N (

cm2 /C

EL

L)

Page 38: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 38

CC Cell

10-13

10-11

10-9

10-7

10-5

0 10 20 30 40 50 60

50 MHz10 MHz

100 MHz

= ALETB

100 MHzA = 3E-13B = 3

50 MHzA = 2E-13B = 3

10 MHzA = 8E-14B = 3

UMC Actel RTSX32/72-SU FPGATAMU Heavy IonsCC CELL Power Law

SRI 9/2005

LET (MeV/(mg/cm2))

CR

OS

S S

EC

TIO

N (

cm2 /C

EL

L)

Page 39: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 39

Cell Sensitivity & Clock Speed

• All cells show CLOCK SPEED (CLK) dependence

• Can be parameterized

10-14

10-13

10-12

10 20 50 100

= A LETB with B = 3

The 'A' parameter can be treated as a function of CLOCK SPEED

C CELL

R CELL

IO BUFFER

CC CELL

UMC Actel RTSX32/72-SU FPGATAMU Heavy IonsSRI 2005

CLOCK SPEED (MHz)

A

CLKCLK

DCA

LETA B

)ln(1

(x) Power LawCLOCK @ B=3

CELL SPEED A C DR 10 6.96E-15

50 3.43E-14 -2.33E+13 6.72E+14100 1.31E-13

C 10 2.82E-1550 1.60E-14 -4.66E+13 1.40E+15100 5.60E-14

IOBUF 10 8.00E-1550 5.10E-14 -2.78E+13 6.06E+14100 7.50E-12

CC 10 8.00E-1450 2.00E-13 1.00E+12 5.08E+13100 3.00E-13

A=1/(C + D*ln(x)/x)

Page 40: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 40

Proton Measurements

• Heavy ion measurements suggested proton SEE sensitivity

• IUCF measurements at 197 MeV• Each sample (S/N 1-6, 8) received a fluence of

1.7E12 particles/cm2 at an average flux of 8.22E8 particles/(cm2 s)– Equivalent to ~100 krad(Si)

• Only IOBUF upsets were observed for protons– Cross section comparable to PROFIT Model predictions

• Propagation delay measured as a function of total ionizing dose

Page 41: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 41

IOBUF Proton Cross Section

10-15

10-14

10-13

1 10 100 1000

PROFIT PredictionRTSX32RTSX72

2.9E-14

B values

cm2/cell

2.0E-14

1.4E-14

9.0E-15

4.0E-15A = 10 MeV

IUCF PROTON DATAUMC Actel RTSX FPGAIO Buffer

SRI 3/2005

Modified 2P Bendel

= B(1-exp(-((E/A)-1)1/2

))4

PROTON ENERGY (MeV)

CR

OS

S S

EC

TIO

N (

cm2 /c

ell)

Page 42: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 42

Proton Propagation Delay (Pd)Measurement

• Measured over a chain of inverters up to 100 krad(Si)

• ΔPd(min) @ 1%

• ΔPd(max) @ 5%

• ΔPd(avg) @ 3%

Device Device # ΔPd (%)

RTSX32 1 4%

RTSX32 2 3%

RTSX32 3 4%

RTSX32 4 2%

RTSX72 5 1%

RTSX72 6 4%

RTSX72 8 5%

Page 43: Radiation Characterization of a Hardened 0.22  m Anti-Fuse Field Programmable Gate Array R.J. Nejad 1, P.A. Rickey 2, K. Konadu 2, W.J. Stapor 2, P.T.

November 2005 43

Summary

• Characterized UMC Actel RT54SX 32 & 72 FPGA devices– Heavy ions & Protons– Operational speeds up to 100 MHz– Extends previous investigations

• Innovative test thread approach– Enables separable cell type cross sections– As-used architecture– Scalable to complex applications

• No destructive SEE observed• SEE cross sections determined for R, C, IO, & CC cells

– IO Buffers are the most sensitive– Clock dependency– All cells (TMR included) have non-zero response to heavy ions

• Proton total dose up to ~100 krad(Si)• Space particle radiation effects can be mitigated through

design