As Semiconductor Devices Shrink so do their Reliability and Lifetimes

National Software and Airborne Electronic Hardware Standardization Conference

August 20-21Denver, CO

Lloyd Condra, BoeingGary Horan, FAA

Bill Scofield, Boeing

Outline

• BackgroundAs semiconductor devices shrink, so do their reliability and lifetimes

• What we have done about itAVSI research results

• What we have yet to doTool development

• Implementation

What Does “COTS” Mean?

32 Flavors of COTS(Baskin Robbins only has 31)

StandardModified System

Equipment

Part

Sub-assembly (module)

Mil-Aero

Industrial

Commercial

Consumer

Our Challenge

We develop processes, methods, and standards that allow our customers to……….

…….Design, produce, certify, and support products using parts and materials from Complex Adaptive Systems.

Our COTS supply chain is a Complex Adaptive System that evolves according to forces beyond our control

The COTS Semiconductor Industry is a Complex Adaptive System

500

400

300

200

100

01994 1996 1998 2000 2002 2004 2006 2008

Voltage Scaling

Cu Conductors

Low-DkDielectrics

Model-based Design

Feat

ure

size

, nm

Driving factors:• Cost• Speed• Size• Time-to-market

‘Incidental’ factors:• Reliability• Configuration continuity• What aerospace needs

A feature is a line width, gate length, etc. of a CMOS gate.

Semiconductor Wearout

1995 2005 2015Year produced

0.1

1.0

10

100

1000

Mean Service life, yrs.

Computer/cell phone lifetimes

Mil/Aero lifetimes

0.5 mm 0.25 mm

130 nm 65 nm 35 nmTechnology

Serv

ice

Life

(yea

rs)

1.0μ

0.1μ

0.35μ0.18μ

1990 1995 2000 2005

10

100

Typical service life goal (10 yrs.)

Margin

Source: E. Snyder (Sandia), IRPS, 2002)

Most microcircuits are designed for 3-10 year

service life

Strong motivation to limit insight into long-

term reliability

Predictions Confirmed by Experience*

*Source: DfR Solutions

Wearout failures (Hot Carrier Injection) in 90nm ASIC devices

• Telecom OEM: 10% failure within 4 years• Process Monitoring OEM: 20% failure within 3 years

Major manufacturer of graphic processor units (GPU) limits maximum junction temperature to 80ºC in order to meet 5-year lifetime requirement

Predictions Confirmed by TestingAVSI #17 Results

Avionics In-service Data

0.25 μm: ~20-50 FIT

90 nm: ~ 100 - 300 FIT

700 FIT

Currentstate-of-the-art

is 45 nm

Test system at Tower semiconductor

What We Have Done About Early Semiconductor Wearout

• Aerospace Vehicle Systems Institute Project #17– Participants: Boeing, Honeywell, Goodrich, GE, Rockwell Collins,

DoD, FAA, NASA– Time span: 2003-2007– Subcontractor: Dr. Joseph Bernstein, U of MD

• Results– Literature search failure mechanisms, models, parameters– Confirmed models by testing– “Alpha” version of FaRBS software– Avionics system design handbook

1.E+01

1.E+02

1.E+03

1.E+02 1.E+03 1.E+04 1.E+05 1.E+06 1.E+07 1.E+08Time (equivalent hours)

FIT

Predictions Confirmed by TestingAVSI #17 Results

Acceptable for

Commercial Applications

Required for

Avionics

20002005

2010

Early Wearout Confirmed!!!

Wearout Mechanisms

Electromigration

Gate oxide

GateSource Drain

Conductingchannel Gate

oxideGateDrain Source

Conducting channel

N-substrate

Trench isolation

P+ P+P-well

N+ N+

Oxide breakdown

Hot carrier injection (HCI)

Negative bias temperature

instability (NBTI)Voltage stresses

Current stresses

e x p ( )T D D B a T D D Bg k T

EV γ− ⋅

ex p ( ) ex p ( )a N B T IgN B T I kT

EVγ− ⋅

ex p ( ) ex p ( )H C D a H C D

d kTE

Vγ

⋅

ex p ( )n a E M

kTEJ − ⋅

TDDB

NBTI

HCI

EM

FaRBS Reliability SoftwareBQR

Reliability

User Supplied

90 nm NBTI DegradationReliability Prediction with FaRBS

β=1

~3 yrs.

~0.98

~2% failure in 2-4 years

0.85

1.00

0.90

0.95

0 205 10 15

Reliability vs Time

Time, (yr)

Rel

iabi

lity,

R(t)

=1-F

(t)

Fast

Slow

Typ

Example FaRBS OutputsReliability and Failure Rate Estimates

36Mb SRAM

0.988

0.99

0.992

0.994

0.996

0.998

1

0 2 4 6 8 10 12 14 16 18 20

Time (years)

Relia

bilit

y

36 MB SRAM90 nm technology 1.2 volts, 70ºC

Board failure rate

0

100

200

300

400

500

600

700

800

900

0 2 4 6 8 10 12 14 16 18 20

Time (years)

FIT

36 MB SRAM 1GB DRAM

Example FaRBS Output

Board failure rate

0

100

200

300

400

500

600

700

800

900

0 2 4 6 8 10 12 14 16 18 20

Time (years)

FIT

Board with one 36Mb SRAM and one 1GB DRAM

1 FIT = 1 failure/109 hrs.

Effect of Feature Size

Source: Wu, E.Y., and R.-P. Vollertson, “On the Weibull Shape Factor of Intrinsic Breakdown of Dielectric Films and Its Accurate Experimental Determination – Part I: Theory, Methodology, Experimental Techniques,” IEEE Transactions on Electronic Devices, vol. 49, no. 12, December 2002. Pp. 2131-2140.

β = 1.3

β = 1.6

β = 2.1β = 4.0 β = 7.3

Effect of Voltage on TDDB

Time-to-fail (s)

100 101 102 103 104

l(l

(1F))

-4

-3

-2

-1

0

1

7.75V7.50V7.25V7.00V

BetaW=1.32

• 256M DRAM, 78nm CMOS process• Gate oxide thickness approximately 5.5nm• Operating voltage 1.5V.• HTOL at 5.0V/125C• Exponential voltage acceleration with γ = 2.7

What We Have Yet To Do

AVSI Project #71– “Commercial” version of FaRBS software (DfR Solutions)– Verify software by test data– “Beta test” FaRBS software tool on selected Boeing systems– Update with future technology data, models, and parameters

Invite participation by others

Provide inputs to aerospace design and reliability documents

– DO-254– MIL-HDBK-217

“Commercial” FaRBS InputsMake AVSI 17 results “user-friendly”

–Graphical user interface (GUI) designed to interact with a wide range of users, e.g., design engineers, reliability engineers, etc.

–Requires a minimal set of inputs• Manufacturer• Manufacturer part number• Duty cycle• Use environment (temperature)

–Assumptions can be modified by expert users• Operation at rated voltage• Only mfr.-specified thermal solutions (no uprating)• International Technology Roadmap for Semiconductors (ITRS)

models and parameters• Applicable to <130nm technology• Default and package failure rates from handbooks (-217,

Telcordia) or part manufacturer

“Commercial” FaRBS OutputsMake AVSI 17 results “user-friendly”

• Failure rate as a function of time• Results can be exported in a .doc or .xls / .csv format• Expert user will be able to extract failure rates for each

failure mechanism• Validation link will

provide details on approach and experimental results

ImplementationUpdate System Reliability and Certification Documents

Hardware Reliability Prediction

MIL-HDBK-217

Hardware Design

Assurance

RTCA DO-254

Software Design

Assurance

RTCA DO-178B

System Certification

Analysis

System FMEA

System FTA

System Functional

Hazard Assessment

Updates needed