Overview of highly accelerated life test (halt)

OVERVIEW OF HIGHLYOVERVIEW OF HIGHLY ACCELERATED LIFEACCELERATED LIFE

TESTChet Haibel

©2011 ASQ & Presentation ChetPresented live on Jan 18th, 2012

http://reliabilitycalendar.org/The_Reliability Calendar/Webinars ‐liability_Calendar/Webinars__English/Webinars_‐_English.html

ASQ Reliability DivisionASQ Reliability Division English Webinar SeriesEnglish Webinar SeriesOne of the monthly webinarsOne of the monthly webinars

on topics of interest to reliability engineers.

To view recorded webinar (available to ASQ Reliability ) /Division members only) visit asq.org/reliability

To sign up for the free and available to anyone live webinars visit reliabilitycalendar.org and select English Webinars to find links to register for upcoming events

http://reliabilitycalendar.org/The_Reliability Calendar/Webinars ‐liability_Calendar/Webinars__English/Webinars_‐_English.html

Chet Haibel ©2012 Hobbs Engineering Corp.

OVERVIEW OF HIGHLY ACCELERATED LIFE TEST

Chet Haibel Hobbs Engineering Corporation

www.hobbsengr.com (303) 465-5988

http://www.hobbsengr.com/

Chet Haibel ©2012 Hobbs Engineering Corp. 1

What Is Reliability?

CLASSICAL DEFINITION

Reliability is the probability that a component,

subassembly, instrument, or system will perform

its specified function for a specified period of time

under specified environmental and use conditions.


What is a Product Failure?

Failure is the inability of a device to perform its intended functions

under stated environmental conditions for a specified time.

Failures are classified into three types based on time:

• Early-Life (Infant Mortality)

• Useful-Life (Random-in-time)

• Wear-Out (End of useful life)

Each failure type has different kinds of causes and therefore different

tests to discover them and different methods of correction / prevention.


What is a Product Failure?

Failures are also classified into three types based on their persistence:

• Hard Failure (Persistent)

Typically a component must be replaced, but trouble-shooting may be

done at room temperature with no vibration or other stimulus

• Soft Failure (Temporary)

Often merely removing the environmental stimulus clears the problem,

but sometimes it is necessary to cycle power, clear fault logs, etc.

Product must be stressed to duplicate and trouble-shoot soft failures

Many very important reliability issues are SOFT FAILURES.

• Intermittent Failure (Elusive)

This is permanent but the failure mode must be put into a detectable state


What Causes Product Failure?

A component fails when applied load exceeds design strength.

Applied Load Design Strength

Failure

Units of Applied Load, Strength


Applied Loads

Examples of applied load might be:

Force

Torque

Tension

Shear

Pressure

Voltage

Current

Wattage

Clock Speed

Electrostatic Discharge

Electromagnetic Interference


Design Strength

Examples of design strength:

Torque rating of a bolt

Voltage rating of a capacitor

Current rating of a diode

Power rating of a resistor

Shear strength of solder

Tensile rating of plastic

Temperature rating of transformer insulation


Load / Strength Interference

Desirable

Obvious

More Subtle

Load Strength

Load Strength

LoadStrength


Early-Life

Useful-Life

Wear-Out

Load Strength

Load Strength

Load Strength

with time

Load / Strength Interference


Bathtub Curve

Operating Time (t)

Haza

rd R

ate

- h

(t)

Useful-Life

Failures

Early-Life

Failures

Wear-Out

Failures

Life to the Beginning of Wear-Out

Random-in-Time Failures


Wear-Out Failures

Hazard

Rate

Time

h(t)Increasing Hazard Rate

Failures due to cycle fatigue

Corrosion

Frictional wear

Shrinkage, cracking in plastic components

Typical of mechanical systems

StrengthLoad

with time


Cycle Fatigue

Cycled by:

• Product Operation

• Thermal Cycling

• Vibration

• Shock

• Etc.

Stresses:

• Pressure

• Tension

• Torsion

• Shear

• Etc.

Use up Fatigue Life


Observed Failure Behavior

For a given stress level, the number of cycles to failure in a sample

will occur in a distribution due to specimen variation

0

2

4

6

8

10

12

14

16

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

Cycles to Failure



Higher stress level requires fewer cycles to failure

0

2

4

6

8

10

12

14

16

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

Cycles to Failure

Higher Stress Lower Stress



For the same failure mode, stress level and the number of cycles

to failure are related by a straight line on log scales

S - N Diagram

0.9

1.0

1.1

1.2

1.3

1.4

1.5

1.6

0 1 2 3 4 5

Log N, Cycles to Failure

Lo

g S

, S

tress

S1 N1

S2 N2


One Failure Mode: Fatigue Damage

Vibration Analysis of Electronic Equipment by Dave Steinberg, Wiley, 1973

D n b, where

• D is the Miner’s Criterion fatigue damage accumulation,

• n is the number of cycles of stress,

• is the stress in force per unit area,

• b is the negative, inverse slope of the S-N diagram for the material.

For wrought Aluminum, doubling the stress decreases the

fatigue cycles by a factor of 1000 b is approximately 10

15


S-N Diagram for 7075 Aluminum

Vibration Analysis of Electronic Equipment by Dave Steinberg, Wiley, 1973

~ 2 thousand cycles at 80 KSI, but at 40 KSI it takes 2 million cycles

O

O


Assume a resonance at 1,000 Hz

At 40,000 psi, failure would occur at 2 million cycles

2 million ÷ 1 kHz = 2000 seconds or 33 minutes

At 80,000 psi, failure would occur in 2 seconds

Doubling the G rms level would achieve a time compression

factor of 1,000.

This TIME COMPRESSION is normal for HALT

Fatigue Damage from Vibration

17


Time Compression

Reference: GE Lighting, private telecon with Jim Harsa in 2000

D t v b

t is time

v is the voltage

b =13 for incandescent lights

b = 8 for fluorescent lights

18

Increased voltage stress shortens

time to see the same dominant

Wear-Out failure mode


Discovering Wear-Out Failures Without Using HALT

If possible, set up a repetitive “cycle test” which removes the “dead

time” between cycles. But brainstorm what test artifact may be

added and / or what the test may be concealing

Test until a minimum of five failures are produced [Haibel’s rule]

Use Weibull Analysis to fit a distribution to the failure data

If life is not sufficient, determine the reservoir of material and the

process consuming the reservoir. Increase the reservoir of material

and / or slow down the process consuming it

If necessary, replace the reservoir of material periodically with a

scheduled preventive maintenance program


Discovering Wear-Out Failures Without Using HALT

Electromigration

(photo courtesy Alcatel-Lucent)

Standard test for

electromigration in

MIL-STD-883 is

Dynamic Burn-In:

125°C for 160 hours

with all voltages,

currents, and clock

speed maximized


Useful-Life Failures

Hazard

Rate

Time

h(t)Constant Hazard Rate

Random-in-time failures

Parts are new until they fail

Strength-Load interference

Insufficient design margin

Typical of electronic hardware

StrengthLoad


Quantifying Strength / Load Interference

2/122 )( LS

LS MMSM

Subtracting two Normal

distributions produces

another Normal

distribution whose mean

is the difference of the

means, but whose

standard deviation is the

root-sum-square of the

two standard deviations

We define Safety Margin


Useful-Life Failures

For simple mechanical products with few parts, we can calculate

reliability one part at a time using Safety Margin for Normal

distributions, or using Monte Carlo simulations for non-Normal

distributions.

For electro-mechanical products with thousands of components

(each of which may have several relevant strength characteristics),

we need an efficient technique to catch the few component

applications that have marginal strength / load relationships. So far,

the most efficient technique is Highly Accelerated Life Test (HALT).

Load Strength


Highly Accelerated Life Test

Used in the Design Phase

HALT

24


HALT Finds Useful-Life Failures

Load Strength

Load Strength

constantly

increasing load

Increase probability of seeing an existing failure mode


HALT is the method of seeing the existing failure modes

with the minimum number of prototypes (4 or 8)

in the minimum time (typically a week)

By experience with early prototypes or with similar

products, determine which environmental factors will

“stimulate” the relevant failure modes

Many failure modes in typical electromechanical

products are well stimulated by temperature and

rapid temperature cycling simultaneous with six

degree-of-freedom random vibration

HALT


G rms

Temperature (Celsius)

20

40

60

80

-20

0

-40

15 20 25 30

ENV2

ENV1

10 5

Goal “limit of technology”

Goal “limit of technology”


HALT

Every stimulus of potential value is used during New Product Development to find the weak links in the product design

These stresses are not meant to simulate field environments but to find the weak links in the design using only a few units in a very short period of time

Stress levels are taken well beyond the normal mission profile

Sometimes one kind of stress will produce a failure mode in HALT, but a different kind of stress will produce that same failure mode in the hands of customers

Focus on fixing the failure mode, don’t focus on the stimulus

Crossover Effect


Crossover Effect


Thermal

Cycle

Vibration

Voltage

Cycle

High Temp

Burn in

All Combined

Margining

Reference: “Flaw-Stimulus Relationships”, G. K. Hobbs, Sound and Vibration, August 1986

Stimulus-Flaw Precipitation Relationships

30


More than one failure mode may be affected by the same stress

Failure modes will not necessarily be exposed according to the field Pareto chart, but maybe in some other order

The time compression factor for the failure modes will be different

Perhaps a Different Order

31

Field

Pareto

HALT

Order


Order of application and discovery:

Cold Step Stress 14%

Hot Step Stress 17%

Temperature Transition 4%

6-Axis Vibration 45%

Combined Temp and Vibe 20%

Without simultaneous, all axis vibration,

65% would have been missed!

“Summary of HALT and HASS Results at an Accelerated Reliability Test Center” by Mike Silverman

Based on 49 products from 19 different industries

Failure % by Stress Type

32


Where Design Flaws Were Discovered

Cold Step Stress 10%

Hot Step Stress 12%

Rapid Thermal Cycling 4%

Vibration Step Stress 43%

Combined Temp and Vibe 31%

74% of the flaws would have been missed

without simultaneous, all axis vibration!

Chuck Laurenson, Parker Hannifin 1999 Hobbs Engineering ARTS USA Award Winning Paper

“Our Path to Reliability Using HALT”

33


Let’s Focus on Vibration Swept Sine, Single Axis

Random, Single Axis

Six Degree of Freedom

34


ElectroDynamic Shaker

35


Z-Axis Mode of Vibration

36


Driven Harmonic Motion

37

Z-axis

excitation

A cos 2πft

ftAzdt

dz

dt

zd2cosKDM

2

2

0.001

0.01

0.1

1

10

1 10 100 1000

Transfer Function

Shaker frequency in Hz


Swept Sine Vibration

Essentially one frequency at a time,

sweeping at one octave per minute

Typically uses a Hydraulic shaker (limited upper frequency) or an

ElectroDynamic shaker (high powered voice coil)

Using a Stroboscope, one can observe behavior at resonance

But can only see one resonance at a time, in one translation

axis at a time; must mount the product for X, Y, & Z

Miss interactions between resonances at different

frequencies or in different directions

No guarantee of stimulating rotational resonances at all !

38


Voice Coil Can be Rotated to Drive the Slip Table for X or Y

39


An Oil Bearing Supports the Slip Table

40


Random Vibration

Broadband, Pseudo Random (noise-like)

vibration generated by a computer

Typically uses an ElectroDynamic shaker, therefore one translation

axis at a time; still have to mount the product three times for X, Y,

& Z and that doesn’t stimulate rotational resonances very well

But this is a major improvement to see all frequencies at once,

therefore see the interaction of resonances in one direction

Crest factor (ratio of peak to average acceleration) is around 3

Major advantage is to shape the spectrum for qualifying to some

external standard (e.g., RCTA/DO-160D Category U Helicopter)

41


Random Vibration Shaped Spectrum

42

0.001

0.010

0.100

1.000

10 100 1000

Po

we

r S

pe

ctr

al D

en

sit

yg

2/H

z

Freqency (Hz)

Vertical axis is

Power Spectral

Density in units

of g2/Hz

To convert to G

rms, integrate

the power (g2)

over frequency

and take the

square root

Shown is approximately 5G rms


TIME COMPRESSORTM TC-1 Ocelot by

HALT & HASS Systems Corporation


44

Temperature change rates of plus or minus 120 Celsius degrees per

minute, the highest in the industry, from -100°C to +200°C

Vibration will start and run anywhere from 0.1 to 150 G rms

Low G levels are important for executing Modulated Excitation™

which is a breakthrough for detecting intermittent failures

X, Y, and Z acceleration balance is near 1:1:1

Sound level is only 50 dBA at 30 G rms, the lowest in the industry,

no ear protection is necessary, can be used on production lines

Will operate on 110 volts, 50-60 Hz with reduced heating for trouble

shooting – this is important for duplicating soft failures

Features of the TC-1 Ocelot


TC-1 Ocelot Vibration System

45

These are

pneumatically-

driven pistons

which generate

six-axis (6 DoF)

vibration from

approximately

20Hz to 10kHz

(one spring is

removed to

show the table

construction

detail) Bottom View


Repetitive Shock Spectrum

46

Time in seconds

T d

Mathematically, a string of

rectangular pulses of period T and

duration d in the Time Domain

Transforms into a “comb” of

frequencies whose fundamental

frequency is 1/T with harmonics

weighted by in the

Frequency Domain πdf

dfSin

0.00001

0.0001

0.001

0.01

0.1

1

Frequency in Hz


Six-Axis Random Vibration

Using several pneumatic pistons, with air flow modulated in a

proprietary fashion, produces overlapping smeared spectrums

The different angles of the pneumatic pistons generate a feedback

controlled, broadband level of random vibration in X, Y, and Z

translational directions and yaw, pitch, and roll angular directions

Feedback for the control system is provided from one z-direction

accelerometer on the bottom (piston side) of the table

This results in all frequencies in all directions, simultaneously

exciting all resonances for complete failure mode stimulus

The Crest Factor, the ratio of peak to average acceleration is ~10,

which rapidly precipitates design and manufacturing flaws

47


Poorly mounted components

Poorly formed leads

Poor solder joints

Fretting Corrosion

Loose hardware

Loose wires

Adjacent parts contacting

Wires over sharp edges

Stacked resonances

Some Defects Precipitated by Vibration

48


Some Defects Precipitated by Vibration

49


Poorly matched expansion coefficients

• Boards and components should match

• Structures should match

Poor solder joints

Improperly formed leads

Improper crimps

PCB shorts, opens

Plated through hole defect

Some Defects Precipitated by Thermal Cycling

50


Some Defects Precipitated by Thermal Cycling

51


Effect of Temperature Rate on Number of Cycles

“Effective and Economics-Yardsticks for ESS Decisions”, S. A. Smithson, IES, 1990

52


Time Compression for Data from the Previous Slide

Calculations by G. K. Hobbs

At a Ramp Rate of 5⁰C per minute, 400 cycles with a range of 165⁰C

(with no dwells) would take 440 hours

At a Ramp Rate of 25⁰C per minute, 4 cycles with a range of 165⁰C

(with no dwells) would take less than 60 minutes

(At a Ramp Rate of 40⁰C per minute, 1 cycle with a range of 165⁰C

(with no dwells) would take less than 10 minutes)

This is real TIME COMPRESSION !


Stresses Used in HALT

Wide range temperature

High rate temp. cycling

All axis random vibration

Power cycling

Power voltage and frequency

Secondary voltage

Digital clock frequency

Humidity

Dimensional parameters

Viscosity of a fluid

Vary pH of a fluid

Salinity of a fluid

Add particulates to the fluid

Back Pressure


More Stresses Used in HALT

Inject electrical noise

Mistune the channel

Radiation (E & M)

Nuclear radiation

Multiple sterilizations

Whatever else makes

sense for the

particular product

Vary magnetic tape thickness

Vary gear diameter

Off axis alignment

Mismatch / Overload

Imbalance

Off-track

Higher RPM


A flaw may be exposed by a different stress in HALT than the stress which exposes the flaw in the field environment

Focus on the failure modes and mechanisms, not the stresses used to expose them or the margin beyond field environment

Focusing on margin may lead to missing an opportunity for improvement followed by field failures of the same mode

This is a frequent, serious mistake in HALT!

Crossover Effect

56


What Level of Stresses to Use


Vibration

• All modes excited

• Second modes are very important

Thermal

• All sites reach the desired temperatures

• All sites reach the desired rates of change

Voltage

Humidity

Current density

Other stresses or parameters

Product Response is of Prime Importance, the Inputs Are Not

58


In HALT, one must go beyond customer-specified stress level to

compress the time to see the dominant failure modes

Stress level has been substituted for sample size!

This is one of the MAJOR BENEFITS of HALT

We do not need many units to HALT (four is good)

We can HALT a few at each stage of development and manufacturing.

• Prototype (as early as feasible)

• Pre-production (after corrections)

• Early production (after design transfer)

• Ongoing production (re-HALT)

59

What Level of Stresses to Use


Understand First

Again, the key is to focus on the failure mode, not the stress type used, or the margin beyond the field environment

Through failure analysis, gain root cause understanding first and then decide if the weakness would cause field failures or whether the weakness would put limitations on manufacturing screening

It’s often easier to fix it than prove it’s not a customer issue!

60


Every weakness found represents an opportunity for improvement

HALT is proactive, but no action means no improvement

We try to break the product in order to find its weak links

This is discovery testing compared to qualification (success) testing

This is a total paradigm shift!

Opportunities not taken will probably lead to field failures much

more expensive than the improvement would have been. This fact

has been documented in thousands of cases

If you find it in HALT, it is probably relevant !

HALT Attitude

61



Boeing 777 was the first

commercial airplane

ever certified for Extended

Twin-engine Operations

(ETOPS) at the outset of

service

Example of Success

Ed Minor, Boeing, in a presentation at a Hobbs Engineering Seminar

63

“Dispatch reliability after only two months of service was

better than the next best commercial airliner after six years”


Accelerometers

Analysis &Test Equipment

ASICs / Processors / Drives

Land / Air / Water Craft

A/V Products & Systems

Avionics / Aerospace

Compressors/Generators

PCs to Mainframes

Lipstick

Electronics / Electrical

Gears / Transmissions

Instruments / Gauges

Magnetic Resonance Scanners

Medical Products

Military / NASA (mixed)

Monitors / Displays / TVs

Ovens

Pneumatic Vibration

Point of Sale Systems

Power Supplies

Radar / GPS Systems

Telecommunications

Thermal Controls

Jet Engines / Missiles

Some Product Types Successfully Improved by HALT


HALT consists of:

• Precipitation

– Stresses

– Stress Levels

• Detection

– Detectable State

– Coverage

• Failure Analysis

• Corrective Action

– Corrective Action Verification

The Complete HALT Process

65

All must be present or no

improvement happens !


Achieve a Detectable State, the “Magic Level” or the “Sweet Spot”

where the intermittent is detectable

• Detection Screens are a well established technique commonly

practiced by the experts

• Requires equipment designed for HALT and HASS for best results

• Modulated ExcitationTM frequently improves detection by two

orders of magnitude, sometimes even more

The First Part of Detection

66


Some damage from the HALT stresses may not be

immediately discernable – it may be LATENT !

HAST (Highly Accelerated Stress Test -- Pressure Cooker) may

precipitate latent damage, making it patent -- discernable

• Cracked component bodies (e.g. MLCC)

• Other long term failure modes not yet completed

If feasible, expose all HALT units to HAST

Or perform a biased (power on with signals toggling) exposure

to 60°C and 90% RH for one week

Detection Excellence

67


Multi-Layer Ceramic Capacitor

CALCE Electronic Products and Systems Center, University of Maryland

PCBA Flexing


Combined all-axis, broad-band vibration and high-rate thermal cycling. Low frequencies must be present in sufficient amplitude to precipitate the defects.

Electrical stressing (power supply, clock frequency, loads)

Monitoring with high coverage is absolutely essential

Temperature, pressure, and humidity (HAST) equipment Traditional 85/85 takes 1,000 to 5,000 hours HAST takes only 48 hours!

Other stressors (such as corrosive atmosphere or radiation) as appropriate for the product and its environments

Equipment Required

69


Appreciating HALT

To Appreciate

HALT, let’s look at

prototype test

quantities required

under normal

conditions


Reasonable Example

Suppose an R&D project has a product reliability

goal to have less than 5% Annual Failure Rate.

(this is not a lofty goal)

How many prototype units would have to be put

on test to have 70% probability of seeing all the

problems that must be resolved to be successful?


Infinite, Decreasing, Geometric Series

Mathematical Model for a Pareto

F1 , F1R , F1R2 , F1R

3 , ...

Sum = F1 / (1 - R) 0 < R < 1

Example:

If sum = 5%, R = 0.8, solve for F1

Answer:

(Sum)(1 - R) = F1 = (5%)(0.2) = 1%


Infinite, Decreasing, Geometric Series

0

1

2

3

4

A B C D E F G H I J K L M N O P

FAILURE MODE

PE

RC

EN

T

“allowed”

failure

modes


10

100

1000

0.001 0.01 0.1

Nu

mb

er o

f u

nit

s on

tes

t

Failure mode's failure probability

0.99 0.90 0.50 0.70

O

70% Chance of Seeing Failures for 5% Annual Failure Rate


Minimum Prototypes and Time

To see the failure modes that must be eliminated

for even mediocre reliability (5% AFR),

Test 120 units for a year at normal mission (customer, field) conditions,

or

HALT 4 units for a week


There are “Accelerated Reliability Test Centers” where you can take

some products to try a HALT chamber

The persons at the ARTC will run the chamber, but you have to

run your product using diagnostic software

Take an existing (currently shipping) product for which you know

the failure modes experienced by your customers

This is an excellent way to prove that HALT will find the relevant

failure modes in YOUR product

How to Prove that HALT Works


HALT WORKSHOP

Preparing to HALT a Product

Preparing a Product for HALT

77



In any test we have to stimulate the product and look for a response

from it. HALT is no different, we need inputs and outputs which

we can control and observe from outside the HALT chamber.

Ideally, we want to check all functions of the product so we can see

any (soft) failures.

We often figure out a “quick test” which we can run at each condition

of voltage, temperature, vibration, etc. This might be the power-on

self-test (POST), so we power cycle the product at each condition.

Then occasionally, we will take the time to do a thorough checkout.

78



Many products (especially software driven products) detect power

supply voltage and will shut down outside an upper and lower limit.

Some products detect temperature and will shut down outside an

upper and lower limit.

These protections must be disabled, either with special HALT

software (firmware) or by modifying the hardware (supplying a

stable voltage to the temperature and / or voltage comparators).

We want to see the underlying (raw) performance of the circuits.

These voltage and temperature limits will improve design margin.

79



Some products have rubber feet on them to reduce skidding and

scratching, and take out minor irregularities in the support surface.

These will tend to dampen the vibration we are trying to drive into the

product. We must overcome this dampening by removing the feet

or supporting the product next to the feet on the chassis.

Similarly, inside the product there may be elastomer material to dampen

vibration. These dampeners must be defeated to transmit vibration.

80



Most products have covers to protect the electronics from foreign

(conductive) material and protect the user from coming in contact

with live voltages.

Some products have fans to circulate air to cool the hot components

(and heat the cool components).

These covers and fans will get in the way of the turbulent airflow in the

HALT chamber, which is trying to impose a temperature on the

components. It makes a convection oven look tame!

Unless these covers are structural, they should be removed. If they are

structural, they must have holes drilled in them to let the airflow in.

81


OVERVIEW OF HIGHLY ACCELERATED LIFE TEST

Chet Haibel Hobbs Engineering Corporation

www.hobbsengr.com (303) 465-5988

http://www.hobbsengr.com/

Overview of highly accelerated life test (halt)

Technology

orgthereliabilitycalendarwebinarsliability

observed failure

time compression

applied load

normal distributions

failure modes

types based

electronic