DPMO ProjectFinal Report

Chair: Tim Kruse, Plexus Corp.Co-Chair: David Mendez, Solectron Corp.

iNEMI Webinar — February 16, 2005

2

Agenda

• Participants

• Background & Motivation

• Scope of Work

• Database Structure

• Sample of Results

• Accomplishments

• Lessons Learned

• Possible Future Activities

3

DPMO Project Participants

• Agilent*

• Celestica*

• Delphi Automotive*

• Georgia Tech

• Hewlett Packard*

• Mack Technologies*

• Motorola

• Nortel

• Plexus*

• Sanmina-SCI*

• Shipley

• Solectron*

• Teradyne

• Universal

• Vitronics-Soltec

* Data Contributors

4

DPMO Definition

• DPMO - Defects per Million Opportunities

• Defined in IPC-7912 and IPC-9261 as number ofdefects divided by number of defect opportunities,multiplied by 106

• Can be categorized many ways:– Process (SMT placement, reflow, wave, etc)

– Defect type (placement, component, termination, assembly)

– Package type

– Others…

5

Background and Motivation

• The NEMI 2002 Roadmap required conversion costreduction in the cost per I/O of each electronic assemblyover time

• A factor in reducing conversion cost is enabling higherquality, thereby:

Reducing manufacturing costs through higher utilization

Shortening product cycle times

Reducing cost of test and inspection

• With data that quantifies the expected defect spectrum on aPCBA, manufactures can accurately assess:

the cost of assembly, test, repair and scrap

the estimated shipped product quality level of a product

6

Background and Motivation

Yield vs. Number of Defect Opportunities

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

0 10,000 20,000 30,000 40,000Defect Opportunities

Est

imat

ed Y

ield

DPMO=10 DPMO=25 DPMO=50 DPMO=100 DPMO=250 DPMO=500 DPMO=1,000

Estimated Yield = e-DPU

DPU = DPMO x # of Ops

IPC 9621

7

Purpose

The project objective was to provide tools that:

– Create an effective environment to continually improvethe delivered quality of manufacturing processes

– Assist in reducing costs of assembly, test, rework, scrapand warranty

– Help improve line utilization and reduce cycle time

– Allow manufacturers to better prioritize the deploymentof constrained resources

– Allow manufacturers to benchmark their DPMO rates toothers in the industry

… regardless of board complexity

8

Scope of Work

• Implement methods to protect confidentiality of data

• Define how data will be shared between teammembers and with outside parties

• Define data collection methods around existing IPCstandards– IPC 9261 In-Process DPMO and Estimated Yield for

PWAs

– IPC 7912 Calculation of DPMO and ManufacturingIndices for PCBAs

9

Scope of Work - continued

• Align efforts with other organizations (IPC &SMART Group)

• Define data stratification and classification methods

• Gather data from several component types intodatabase

• Summarize data and generate reports anddocuments

• Lay foundation for ongoing DPMO metric efforts

10

Data Sharing Policies

• Data shared with the industry

Methodology, database format, & data classifications

Statistics detailing the size and content of the database

DPMO level for all board types averaged together withineach company

• Package-level DPMO data will not be shared withthe industry

• Project participants receive the entire database

11

Confidentiality of Data

• Established a secure website for project information

• Participants submitted data to a neutral party in thepre-defined formats

• The neutral party replaced the company nameswith non-descript IDs

• The data was combined into a single database

• The combined database was distributed to projectparticipants

• Georgia Tech was the neutral party

12

Database Structure

13

Defect Types

Defect Types• Agreed to 22 defects categories• Where possible, aligned NEMI defect categories to

other industry initiatives:• The SMART Group’s PPM project• Classified defect categories per IPC Defect

Classification for DPMO calculations (Appendix A of IPC9261)

• Defect was defined as anomaly that required arepair/rework action.

• False calls, process indicators, or “self-correctingdefects” that are not reworked/repaired were notlogged.

14

Defect Types - continued

TerminationOTHER TERMINATION DEFECT

TerminationSOLDER BALL

TerminationSOLDER TERMINATION SHAPE (excess/void/other)

TerminationSOLDER INSUFFICENT

TerminationSOLDER TERMINATION OPEN

TerminationSOLDER TERMINATION BRIDGE - SHORT

AssemblyOTHER PASTE DEFECT

AssemblyPASTE SCOOPING

AssemblyPASTE BRIDGING

AssemblyPASTE SMEARING

AssemblyPASTE INSUFFICIENT

IPC DefectClassification

NEMI DPMO Project

Defect Name

15

Defect Types - continued

AssemblyOTHER DEFECT

ComponentOTHER COMPONENT DEFECT (electrical /mechanical)

ComponentBARE BOARD DEFECT (trace/warp)

ComponentMECHANICAL ASSEMBLY DEFECT (heatsink,screw…)

Component*COMPONENT DAMAGED

ComponentCOMPONENT ELECTRICALLY DEFECTIVE

ComponentCOMPONENT LEAD BENT or MISSING LEAD

PlacementCOMPONENT PLACEMENT (misaligned, billboard…)

PlacementCOMPONENT MISSING

PlacementCOMPONENT ORIENTATION (90/180/270)

PlacementCOMPONENT WRONG

IPC DefectClassification

NEMI DPMO Project

Defect Name

16

Package Types

Package Types

• Agreed to 36 package type categories based oninput from several participants

• Targeted a list that would cover most packagetypes while maintaining a significant number ofopportunities per package type.

17

Package Types - continued

Gullwing connectors, regardless of pitch, (does not include Mictor or Straddle Mount)GW CONN

greater than 25 mill gull wing components and hardware (sockets, switches), SOT - SmallOutline Transitors, SOD, DPAKsGW GT 25 MIL

25 mil gull wing QFP, SOIC, SOP, SSOP, TSOP, TSSOP and hardware (sockets, switches)GW 25 MIL

20 mil gull wing QFP, SOIC, SOP, SSOP, TSOP, TSSOP and hardware (sockets, switches)GW 20 MIL

16 mil gull wing QFP, SOIC, SOP, SSOP, TSOP, TSSOP and hardware (sockets, switches),also includes 12 mil gull wingGW 16 MIL

Ball Grid Array. Standard pitch is 1mm (or 0.039"). Includes plastic, ceramic, eutectic, high-melt.PGA

FCA - Flip Chip Array - a chip on board technology that has bumps attached to the silicondie, is flipped, and mounted directly to a printed wiring board.FLIP CHIP ARRAY

Column Grid Array, fine pitch, Anything less than standard BGA pitch (1mm or 0.039") mayrequire special handling.CGA FP

Column Grid Array, similar to BGA except terminations are small columns of solderCGA

Includes any BGA connector regardless of pitch.BGA CONN

BGA - Fine Pitch. Anything less than standard BGA pitch (1mm or 0.039") may requirespecial handling. Includes CSPsBGA FP

Ball Grid Array. Standard pitch is 1mm (or 0.039"). Includes plastic, ceramic, eutectic, high-melt.BGA

Package DescriptionPackage Name

18

Package Types - continued

Fiber Optic cables and componentsOPTICA SMT component that falls into no other category (EMI shields, ground planes,inductors, MELF components, non-leaded )SMT MISC

Press fit connectors and components that are hand pluged, includes ICs hand placedinto chip carriersPRESS FIT

Consider 1 componentMULTICHIP MODULE

Edge connector with SMT leads on both sides of the board, Includes StraddlemountMictors

STRADDLEMOUNTCONN

Mictor connector has both SMT and PTH pinsMICTOR CONN

Bolt, clip, kit, nut, pin, rivet, screw, spacer, spring, standoff, washer,MECH FASTENER

Bearing, chassis, faceplates, fan guard, fuse holder, handle, heatsink, plastic part, RFsheilds,MECH ASSEM

LCC- Leadless Chip Carrier, also LCCC- leadless ceramic chip carrier, both havesolderable castellations for terminationsLCC

land grid arrayLAND GRID ARRAY

LabelsLABEL

SOJ - Small Outline J-leads, PLCC - Plastic Leaded Chip Carriers, socketsJ LEAD

Package DescriptionPackage Name

19

Package Types - continued

Tantalum capacitorsTANT

Wire additions and/or trace cuts due to design modificationsWIRE ADDS / CUTS

greater than 0805 chip components (excluding Tantalum)GT 0805

0805 chip components0805

0603 chip components0603

0402 chip components0402

0201 chip components0201

Resistor networks, capacitor networks, RPACK, (only if device has solderablecastellations, if leaded falls into corresponding category (PTH COMP, GW, etc)

SMT PASSIVENETWORKS

Printed Wiring (or PCB-Circuit) Board, includes gold fingers, buried capacitors/resistorsPWB

Pin Through Hole Array, connectors, sockets - 50 mil and belowPTH CONN FP

Pin Through Hole Array, connectors, sockets - above 50 milPTH CONN

Pin Through Hole Componenet, SIP, DIP, radial and axial mount components, Alsojumpers,PTH COMP

Package DescriptionPackage Name

20

62 10¥

˙˙˚

˘

ÍÍÎ

È

+++

+++=

   Â   Â

atpc

atpc

oooo

ddddIndexDPMO

61 10¥

˙˙˚

˘

ÍÍÎ

È

++

++=

  Â  Â

tpc

tpc

ooo

dddIndexDPMO

DPMO Calculations

dc = # of component defects

dp = # of placement defects

dt = # of termination defects

da = # of assembly defects

oc = # of component opportunities

op = # of placement opportunities

ot = # of termination opportunities

oa = # of assembly opportunities

Example: Component DPMO = DPMOc = [ ∑dc / ∑oc ] x 106

21

Database Statistics - Board Types

335,467Number of boards

42,967

2

883

Maximum volume per board type

Minimum volume per board type

Average volume per board type

380# of different board types

8 / 11Data contributing companies / mfg sites

22

Database Statistics - DPMO Levels

123380,905,70546,941Component

3,215,873,704

335,467

380,571,642

2,454,060,890

Opportunities

9,0263,028Assembly

342,258

93,414

198,875

Defects

81Termination

245Placement

106Total

DPMO

(dx / ox) x 106

Defects & Ops inthe Database

23

DPMO by Company

154

147

140

118

98

90

57

34

24

16

7

DPMOIndex2

30

12

5

292

2

23

7

5

2

1

1

# ofAssys

50415053,41111427032811

1,9641456,321705784910

38013318,888131186879

5971176,7421032101118

5,411971,109322741057

6778327,564192971996

6,642564,968163551095

6533257,2575741664

3512401782193

42,9671605150572

6187071141

Avg.Volume

DPMOIndex1DPMOaDPMOtDPMOpDPMOcCompany

24

DPMO by Mfg Process

0

50

100

150

200

250

300

350

400

Reflow

- SM

T Han

d Plac

ed

Reflow

- SM

T Mac

hine

Placed

Reflow

- PTH H

and

Placed

Wav

e - P

TH Han

d Plac

ed

Wav

e - P

TH Mac

hine

Placed

Wav

e - S

MT M

achin

e Plac

ed

Hand

Solder

- SM

T Han

d Plac

ed

Hand

Solder

- PTH H

and

Placed

Mec

hanic

al Ass

embly

Press

Fit

Mfg Process

DP

MO

1

10

100

1,000

10,000

100,000

1,000,000

10,000,000

100,000,000

1,000,000,000

10,000,000,000

Op

po

rtu

nit

ies

(Lo

g S

cale

)

Opportunities

DMPO

1M Ops

1,510

25

DPMO - Discrete Package Types

0

2040

60

80

100120

140

160

TANT

SMTPASSIV

ENETWORK

0402

GT080

5

0805

0603

Package

DP

MO

01002003004005006007008009001,000

Mil

lio

ns

Op

po

rtu

nit

ies

Opportunities

DPMO

26

DPMO - ICs (Leaded Perimeter)

0

50

100

150

200

250

GW20MIL JLEAD GW25MIL GW16MIL GWGT25MIL

Package

DP

MO

0

50

100

150

200

250

300

350

400

450

500

Mil

lio

ns

Op

po

rtu

nit

ies

OpportunitiesDPMO

27

DPMO - ICs (Area Array)

050

100150200250300350400

PGACGA

BGA

BGAFP

FLIP

CHIPARRAY

LANDGRID

ARRAY

Package

DP

MO

0100200300

400500600700

Mil

lio

ns

Op

po

rtu

nit

ies

Opportunities

DPMO

28

DPMO - Connectors

0

100

200

300

400

500

600

GWCONN

STRAD

DLEMOUNTC

ONN

PTHCONN

PRES

SFIT

Package

DP

MO

0

2040

6080

100

120140

Mil

lio

ns

Op

po

rtu

nit

ies

Opportunities

DPMO

29

Pckg Breakdown - Defect Spectrum

Breakdown of Defects for 0603 Package TypeMfg Process = Machine Placed, Reflow Solder

1%1%2%2%3%3%3%

9%

17%

40%

19%

0

5

10

15

20

25

30

COMPONENTM

ISSIN

G

SOLDERTERM

INATIO

NOPEN

COMPONENTPLA

CEMENT

COMPONENTW

RONG

COMPONENTORIE

NTATION

COMPONENTDAM

AGED

SOLDERTERM

INATIO

NBRIDGESHORT

SOLDERIN

SUFFICIE

NT

COMPONENTELE

CTRICALL

YDEFECTIVE

SOLDERTERM

INATIO

NSHAPE

OTHER

Th

ou

san

ds

Nu

mb

er o

f D

efec

ts

0%

5%

10%

15%

20%

25%

30%

35%

40%

45%

Per

cen

t o

f T

ota

l Def

ects

30

Pckg Breakdown – 0603

Of the 36 possible package types, 20 had “significant” data• 1,000,000+ opportunities and

• Used on 10+ different assemblies

TANTSMT PASSIVE NETWORKSMECH FASTENERGW 25 MIL

GT 0805PTH CONNMECH ASSEMGW 20 MIL

0805PTH COMPJ LEADGW 16 MIL

0603PRESS FITGW CONNCGA

0402SMT MISCGW GT 25 MILBGA

6665,020980,129,286Overall

487145,640Machine Placed, Wave Soldered

6665,013979,983,646Machine Placed, Reflow Soldered

Package DPMODefectsOpportunitiesMfg Operation

Package = 0603

31

Pckg Breakdown - Statistical Analysis

9007506004503001500

Median

Mean

100806040

Anderson-Darling Normality Test

Variance 19297.494Skewness 3.0758Kurtosis 10.6498N 336

Minimum 0.000

A-Squared

1st Quartile 25.000Median 48.5003rd Quartile 95.000Maximum 944.000

95% Confidence I nterval for Mean79.899

42.68

109.71495% Confidence I nterval for Median

43.000 56.00095% Confidence I nterval for StDev

129.146 150.296

P-Value < 0.005

Mean 94.807StDev 138.915

95% Confidence I ntervals

Analysis of 0603 package (DPMO per assembly)

32

Accomplishments

Met project purpose and scope:• Database with significant defect opportunities for 20

package types and 8 processes across 8 companies

• Team members feel data will be useful in supporting:

– Quality improvement• DFX efforts / component selection

• Applying test and inspection where needed

• Identify potential problem areas / justify necessary changes

– Cost estimation/reduction• Improved DPMO/Yield estimates, conversion cost estimates

• Throughput improvements through better application of constrainedresources

– Benchmarking

33

Challenges & Concerns

• Significant challenges in gathering DPMO data by package!!!

– Data often is stored in different systems

– Few companies have systems linked to automatically compile the data

– Existing IPC specs don’t address package-DPMO

– New releases of IPC specs require even more data to be collected(defects and opportunities on non-populated locations)

– Data collection challenges extended project timeline ~ 12 months

• Caution must be used when interpreting data

– Cannot perform blind comparisons based solely on DPMO

– Effectiveness of test/inspection steps can significantly impact DPMO

– Despite “sanity checks” on the submitted data, data accuracy isultimately dependent on entry points

– A method for normalizing such factors would be useful

34

Possible Follow-on Activities

• Develop a standard for Package-Level DPMO• Use DPMO data in developing “Complexity” or

“Manufacturability” Index• Repeat project / update database at regular intervals• Potentially modify the database structure in support of

follow-on projects• Component specific DPMO (mfg name and part number)

• Inclusion of process parametric data (chemistry, equipment type, ...)

• Scale back data fields to ease data gathering challenges

• Potentially modify defect/package categories based on DPMOdistributions

• Assess impact of process variables (lead free) to DPMO

35

More Information?

• Project information available to the public is at:http://www.nemi.org/projects/ba/dpmo.html

– Database architecture and definition

– Empty database for download and use

– Data-entry guideline with package type, defect types,…

– Industry presentations

– Reports and White Papers as published

• For more information contact project chair/co-chair

– Tim Kruse (Tim.Kruse@plexus.com)

– Dave Mendez (DavidMendez@tx.slr.com)

36

www.inemi.orgwww.inemi.orgEmail contacts:Email contacts:

Tim KruseTim Krusetim.kruse@plexus.comtim.kruse@plexus.com

David MendezDavid Mendezdavidmendez@tx.slr.comdavidmendez@tx.slr.com