Reducing Uncertainty

1NIST/GIPWoG - Fred ByersOctober 14, 2004

Reducing Uncertainty

National Institute of Standards and Technology (NIST)

Information Technology LaboratoryInformation Access Division

Fred ByersNIST

100 Bureau Drive, Stop 8940Gaithersburg, MD 20899Email: [email protected]

October 14, 2004 NIST/GIPWoG - Fred Byers 2

Topics

Consumer ViewLife Expectancy Test IssuesAlternatives Benefits


Manufacturers vs. Headlines

Manufacturers

• Life Expectancy Best Case– ~ 100 – 300 years CD-Rs

– ~ 60 – 100+ years DVD-Rs

Headlines

• Life Expectancy Worst Case– CDs fail in short time frame

– 2 -10 years failure

– “CD Rot”

Consumer View


Most Commonly Asked Question: How long will my (recordable) disc last?

(CD-R, DVD-R, DVD+R)

Answer: It depends on:• Care and handling

– Physical handling

– Environmental influences

• Initial recording quality– Condition of disc before recording

– Quality of burn (depends on disc and burner)

• Disc construction– Materials

– Manufacturing process

Consumer View


More Consumer Questions:

• Should I trust my digital content for long-term storage using these discs?

• Which disc should I buy?• What is the minimum number of years I can expect?• Does price make a difference?• Does brand make a difference?• What should I look for in a disc?• Does the recorder make a difference?

These above points are not unique to optical discs.

Consumer View


What can the user control?

Proper handling– No scratches, etc.

Proper storage– Controlled at recommended storage conditions

• ? Initial recording quality– Good disc – clean, defect free, good recorder

Disc construction – Dependant on manufacturer

Consumer View


- Relatively New -Long-Term Quality Discs

Archive, or long-term labeled discs are available from some manufacturers

• What does it mean? – How are they different from regular discs?– How many years will they last? Longer LE?– How does manufacturer A compare to manufacturer B?– How are they determined to be longer lasting?

Consumer View


Disc-Test Proposal

• A test to measure for minimum longevity expectation

• A test that can be performed in a reasonable time for manufacturers

• An indicator to consumers about minimum quality

• A minimum longevity that consumers can use for planning

Consumer View


Why

• Variation in life expectancy among discs – Quality differences between brands

– Quality differences within brands

• General claims of up to 100 years LE (DVDR)– Actual could be as low as 30 years in normal ambient room conditions

– Some anecdotal stories conflicting with general claims

• User confidence/uncertainty/awareness– Consumer expectations - uncertain or unrealistic

– Planning under uncertainty – refreshing, sampling and migration

Consumer View


PIE fromMetal-Halide Light

0

500

1000

1500

2000

0 100 200 300 400 500Hours

PI e

rro

r

Aged in high temperature and humidity conditions

DVD-R discs

0

500

1000

1500

2000

0 200 400 600 800

D1

D2

D3

HoursP

I err

or

Aged in Metal-Halide light

DVD-R discs

PIE fromHigh T and RH

DVD-R Accelerated AgingComparison Examples

Consumer View


Life Expectancy (LE) Test Standards

Existing LE Test Method Standards

• CD-ROM– ISO 18925:2002, AES 28-1997, ANSI/NAPM IT9.21

• CD-R– ISO 18927:2002, AES 38-2000

Consumer View


LE Testing Issues

• Time duration– Typically one to two years

– Time to market

• Capability– Equipment

– Space

– Expertise

• Cost– Labor

– Contract out

• Existing LE Test Followed?

• No Standard DVD Test

– No Standard LE Test for DVD yet

– NIST is proposing a DVD procedure following the CD Standard

• No Standard “archive quality” test methodology

– NIST proposing a model

LE Issues


LE - Accelerated Aging Times Stress

Test-SetStressed at (Tinc, RHinc)

Incubation duration

Minimum total time

Specimenquantity

1 80 ºC, 85 % 500 h 2000 h 10

2 80 ºC, 70 % 500 h 2000 h 10

3 80 ºC, 55 % 500 h 2000 h 15

4 70 ºC, 85 % 750 h 3000 h 15

5 60 ºC, 85 % 1000 h 4000 h 30

13,000 h total

80total

LE Issues


Incubation + Testing TimeNo. of

ChambersIncubation

timeNo. of weeks

Testing time (2 analyzers)

TotalTime

1 13,000 hrs 78 wks 6 wks 84 wks

2 7,000 hrs 42 wks 6 wks 48 wks

3 5,000 hrs 30 wks 6 wks 36 wks

4 4,000 hrs 24 wks 6 wks 30 wks

LE Issues


Alternatives to the Existing LE Test (to save time and cost)

• Shorten the LE Test– Less expensive

but

– Less accurate

• Initial Error Rate Test– Measures initial quality of data on disc

but

– Does not account for disc degradation rate

– Does not indicate life expectancy

• Target Test - “Archival”, or “Longevity” or “Grade”– Longevity specific, i.e.: minimum number of years expected

– Not a test to determine total disc LE

Alternatives


Alternatives Continued

• Shorten the LE test– Creates higher uncertainty for total disc LE measurement

But

– Uncertainty level should still be acceptable for a lower limit

• Consider error increase-rate– > x = fail

• Consider initial measurement?– BLER-max > x = not acceptable

– Reflectance < x = not acceptable

Alternatives


Another ApproachDon’t calculate for LE (T50) but make it a known lower limit.

• Example:

Make LE (or T50) = 50 years = lower limit.

(At storage conditions 20ºC, 50%RH)

Determine from stress tests if disc will perform beyond the lower limit.

Alternatives


Existing Equation from CD StandardExisting equation from CD-R Standard:T50 = AeH/kTe(B)RH or lnT50 = lnA + H/kT + (B)RH

T50 = Time duration for median disc in a test-set to reach BLER-max (220).A, H and B determines the rate at which median disc reaches maximum allowed errors (220).

Normally we derive A, H and B from test data, then calculate for T50.

Constants:

k

T (Temperature)

RH

Derive from test data:

A

H

B

Calculate:

T50 (Time)

Alternatives


Alternative Use of Existing EquationUsing T50 = AeH/kTe(B)RH or lnT50 = lnA + H/kT + (B)RH

• Instead of calculating for T50, make T50 known.

• For example: T50 50 years (disc reaches BLER-max of 220 in 50 years at a given T and RH)

• Find A and combined e coefficients (H and B) for T50= 50yrs

Constants:

k

T (Temperature)

RH

T50

Derive from T50= 50:

A

H

B

Do not Calculate T50 :

Make T50 = 50years

Use:

lnT50 = lnA + H/kT+ (B)RH lnT50 = 12.99 at 20ºC, 50%RH

Alternatives


First Stress Test Set500 hour intervals x 4

0

100

200

300

400

500

600Y

1 2 3 4 5

Rows

80C, 85%RH


Targeted Time Periods?(“Sunscreen” Approach)

• Can have more than one target or establish only one target.– Total disc LE is unknown (not necessary)

– > 30 yrs

– > 50 yrs

– > 75 yrs

– > 100 yrs

• A test for one targeted minimum longevity time period– 50 year example

– Discs are expected to extrapolate to beyond the lower-limit threshold.

– Error rate increases less than an established maximum acceptability

Alternatives


What will this do for the Consumer? Consumer Indecision Revisited:

• Should I trust my digital content for long-term storage using these discs? (Properly tested discs, yes)

• Which disc should I buy? (Tested discs)

• What is the minimum number of years that I can expect? (As labeled)

• Will price make a difference? (Maybe)

• Will brand make a difference? (Maybe: quality control consistency)

• What should I look for in a disc? (Label showing that the disc batch has passed the test)

• Does the recorder make a difference? (Maybe, same as before)

Consumer View


Consumer/Industry Benefits• Consumer

– Consumer uncertainty reduced

– More informed choice for consumer

– Increase consumer confidence

– More realistic consumer expectation

– Consumer purchasing can be based on needs or migration plans

• Industry– Industry-wide standard test

– Self-test or third party

– Time to market

– Testing cost (compared to existing LE testing)

– Pricing (cost recovery for implementing new procedure)

Benefits


What Is Needed From You, The User

• Define what “archival” means to you or what length of time needed for long-term storage.

• What is the minimum length of time needed for a disc to last to satisfy your requirements?

• What is the ideal length of time?

• Will this make a difference in your purchasing decision?

• Will you look for these “labeled” discs for long-term storage applications instead of other discs?

Consumer View


Other ideas

• Early warning indicator

• RFID


Early Warning Indicator

• Could also be called:– Error Alert

– Check Disc

• A warning about error rates that are approaching BLER max or PIE max.– May also consider Burst errors

• Just a light as an indicator or a pop-up window

• Gives the actual number (good for initial error rate)


Thank you!NIST

Information Access Division

Information Technology LaboratoryDigital Data Preservation

Fred Byers, Oliver Slattery, Jian Zheng

http://www.itl.nist.gov/div895/preservation/

Care and Handling Guide

http://www.itl.nist.gov/div895/carefordisc/


Second Stress Test Set 500 hour intervals x 4

LE Issues

0

100

200

300

400

500

600Y

1 2 3 4 5

Rows

80C, 70%RH


Third Stress Test Set 500 hour intervals x 4

LE Issues

0

100

200

300

400

500

600Y

1 2 3 4 5

Rows

80C, 55%RH


Fourth Stress Test Set 750 hour intervals x 4

LE Issues

0

100

200

300

400

500

600Y

1 2 3 4 5

Rows

70C, 85%RH


Fifth Stress Test Set 1000 hour intervals x 4

LE Issues

0

100

200

300

400

500

600Y

1 2 3 4 5

Rows

60C, 85%RH

Reducing Uncertainty

Documents

Transcript of Reducing Uncertainty