accelerated life testing - sandia...

Inverter Reliability Workshop

Accelerated Life Testing

Rob SorensenSandia National [email protected]

Sandia is a multiprogram laboratory operated by Sandia Corporation, a Lockheed Martin Company, for the United States Department of Energy’s National Nuclear Security Administration under contract DE-AC04-94AL85000This work was performed under funding from the DOE Solar Energy Technologies Program.


• No PV specific industry standard exists• HALT testing is spotty; independently applied• Separate needs identified for residential and commercial scale inverters • Failure modes identified but not in a uniform program applicable across the

industry • System predictive models will require inputs for inverters

Accelerated Aging for Inverters


Laboratory testing provides vital information for PV system reliability

Field Data (O&M, Failures, …)

Accelerated Testing / Lab Tests

System performance model must include wear out (end of life) information

Accelerated Aging of Tape Joint – Thermal

Cycling

ALT

Acceleration Factors

Lin k to Performance

http://upload.wikimedia.org/wikipedia/commons/6/6e/Bathtub_curve.jpg�


What? Component life tests High stresses

• Single or combined• Activate “appropriate” failure modes• Measureable

Failure analysisWhy? Time Full system is expensive and complicated

What is ALT & why?


‘Accelerated Aging’

5

Tem

pera

ture

Reaction Coordinate

25°C, 45 days

40°C, 30 days

100°C, 12 min

500°C, 5 min


6

High T data are extrapolated to “use” conditions (room temperature)


7

How you extrapolate can influence lifetime predictions.

~50% in ~ 1000 years

~50% in ~ 100 years

~50% in ~ 20 years


Two approaches to accelerated testing are used throughout industry

Qualitative Accelerated Tests• HALT tests• HAST tests• HASS tests

Quantitative Accelerated Life Tests• Controlled application of accelerated

stress• Produces acceleration factors (AF) Usage rate acceleration

(Time compression) Overstress acceleration

Small sample sizeSevere level of stress

Increase reliability (product improvement)Qualify new designsDesign quantitative ALT

Reliability under normal use conditions

Used to determine TTFDetermine reliability

Long TimeNeed degradation / failure mechanisms


The Goal of an ALT program is to produce acceleration factors

Often empirical correlations

Limited root-cause analyses

++

−=

obVabVa

n

oRHRH

T1

oT1

kaE

expAF

=

testMTTFMTTFAF field

Time-to-Fail40

1

510

25

50

75

9095

99

50 80 100Cu

mul

ativ

eFa

ilure

Perc

entil

e

MTTF


Empirical relationships may not cut it!

10080604020

exp [B(T + RH)](RH) exp (E /kT)a

n

exp [E /kT + B(RH) ]a2

exp [E /kT + B/(RH)]a

exp [A/kT + B(RH)/kT + C(RH)]

RH (%)

AF

0

1

102

104

106

Calibration data collected here

Atmospheric Corrosion of Micrelectronics

Five accepted environmental models.All agree at 85% RH (ALT conditions).Orders of magnitude difference at 30% RH (use conditions).

ALT must capture valid degradation / failure mechanisms


Issues with ALT

Unknown failure mechanisms Unknown / variable use environment Changing mechanisms as function

of environmental stress Difficult to control and

characterize defects Long duration experiments Evolving / improving technology


What are the likely stresses that lead to Inverter Failure?

Voltage Temperature Thermal cycling Thermal Shock Vibration Mechanical Shock Humidity ??? ??? ???

Heat

Light Moisture

Other


How do we apply ALT to predicting end-of-life (wear out)?

Determine Failure Mechanisms (field exposures)

Understands stresses that activate the failure mechanisms

Devise ALT conditions that accelerate degradation / failure

Perform ALT (range of conditions)

Determine acceleration factors

Qualify Product

Improve Product Reliability

Validate New Designs

Estimate or Predict Reliabilty


Example: Al bondpad corrosion: corrosion requires moisture and contamination & is accelerated by temperature.

14

0

500

1000

1500

2000

2500

3000

0 1000 2000 3000

Res

ista

nce

(m-o

hm)

Exposure Time (hrs)

1ug/cm2, 30C, 80% RHPEM 402

d1-w 1

d1-w 3

d1-w 4

d1-w 11

d1-w 12

d1-w 14

d1-w 15

d1-w 22

Three environmental variables (T, RH, [Cl])

No Failures!


Increasing contaminant level causes failure.

15

0

500

1000

1500

2000

2500

3000

0 50 100 150 200

Res

ista

nce

(m-o

hm)

Exposure Time (hrs)

20 ug/cm2, 30C, 80% RHPEM 414

d1-w 1

d1-w 3

d1-w 4

d1-w 11

d1-w 12

d1-w 14

d1-w 15

d1-w 22

0

500

1000

1500

2000

2500

3000

0 1 2 3

Res

ista

nce

(m-o

hm)

Exposure Time (hrs)

43 ug/cm3, 30C, 80% RHPEM 416

d1-w 1

d1-w 3

d1-w 4

d1-w 11

d1-w 12

d1-w 14

d1-w 15

d1-w 22

20X chloride

40X chloride

Distribution of failure timesNot all failedClear effect of [Cl]


Statistical treatment (life-data analysis) provides a means of analyzing the bondpad data

16

ReliaSoft Weibull++ 7 - www.ReliaSoft.comReliability vs Time Plot

β=0.2360, η=943.6290, ρ=0.8798

Time, (t)

Relia

bility

, R(t)

=1-F(

t)

0.000 1000.000200.000 400.000 600.000 800.0000.000

1.000

0.200

0.400

0.600

0.800

x 4

x 3

x 2

x 3

Reliability

Data 1Weibull-2PRRX SRM MED FMF=16/S=15

Data PointsReliability Line

Rob SorensenSandia National Labs5/28/201010:46:25 AM

10 ug/cm2, 30C, 30% RH


β=0.7073, η=5.2052, ρ=0.9889

Time, (t)

Relia

bility

, R(t)

=1-F(

t)

0.000 90.00018.000 36.000 54.000 72.0000.000

1.000

0.200

0.400

0.600

0.800

x 2

Reliability




20 ug/cm2, 30C, 80% RH


β=2.0355, η=0.6531, γ=0.3818, ρ=0.9576

Time, (t)

Relia

bility

, R(t)

=1-F(

t)

0.000 3.0000.600 1.200 1.800 2.4000.000

1.000

0.200

0.400

0.600

0.800

x 2

x 2

x 3

x 2

Reliability




40 ug/cm2, 30C, 80%RH

• Provides distributions• Includes suspension results• Basis for models {Pfail = f(T, RH, [Cl])}

=

testMTTFMTTFAF field


Bondpad corrosion is modeled as an additional series resistor

Voltage Comparator Circuit Prediction of the effect of corrosion on

LM185 reliability as f(storage location)

The distributed wirebond property (probability of failure) is input into an electrical system model & other component outcomes (reliability, performance threshold) can be determined

y p

0.97

0.975

0.98

0.985

0.99

0.995

1

0 20 40 60 80 100Time (years)

Desert

Gulf Coast

Arctic


Laboratory testing provides vital information for PV system reliability

Field Data (O&M, Failures, …)

Accelerated Testing / Lab Tests

System performance model must include wear out (end of life) information

Accelerated Aging of Tape Joint – Thermal

Cycling

ALT

Acceleration Factors

Lin k to Performance

http://upload.wikimedia.org/wikipedia/commons/6/6e/Bathtub_curve.jpg�


Example: Conductive metal foil tape (Module)

))(028.0(10 α+= tR

Generate ALT data

Develop “acceleration factors”

Module

Load

RIE ×=RIVIP ×=×= 2

Mean

+2σ

-2σ

Determine performance effect

Apply acceleration factors to field


Use the ALT data to predict long-term performance degradation (wear-out???)

20

accelerated life testing - sandia...

Documents