Conductive Anodic Filament

(CAF) Formation in Printed

Circuit Boards (PCBs)
…And What Sherlock Can Do For You
October 11, 2012

1
What is Conductive Anodic Filament (CAF)
Formation?

2
This is CAF…

3
…and this…
Z

A B

Z Z:Z Cross-Section
The photo in A shows a cross-sectional view of conductive anodic filaments
between two plated through holes (PTHs). An oblique slice through the copper
filament is shown in B.

4
…and this.

5
Definitions of Conductive Anodic Filament (CAF)

o IPC-9691 Users Guide for the IPC-TM-650, Method 2.6.25

o A growth consisting of a conductive copper-containing salt. It is
created electrochemically and grows from the anode toward
the cathode subsurface along the epoxy/glass interface

o IPC-TM-650, Method 2.6.25 Conductive Anodic Filament

(CAF) Resistance Test: X-Y Axis
o A form of electrochemical migration within a printed wiring
board (PWB)

o What is electrochemical migration?

6
Definition of Electrochemical Migration

o IPC-TR-476A, Electrochemical Migration: Electrically

Induced Failures in Printed Wiring Assemblies
o The growth of conductive metal filaments or dendrites on or
through a printed board under the influence of a DC voltage
bias

o DfR Solutions
o Movement of metal through an electrolytic solution under an
applied electric field between insulated conductors

7
Schematic of CAF
DC voltage source Electrolytic Cell

e-

_
+

Cathode
Anode

Cations

Electrolyte
Anode: the positive electrode of an electrochemical cell at which oxidation occurs

8
 Why do I care about CAF?

9
CAF causes failures…

10
…and more failures…

11
…and even more failures (typically with burning)

The copper seen here along the outside of these fibers

caused a short in this PCB
12
 How Does CAF Occur?

13
The Four Steps of ECM

o Traditional electrochemical
migration involves four steps
o Path Formation
o Electrodissolution
o Ion Migration
o Electrodeposition

14
Path Formation

o Hollow Fibers

o Drilling Damage

o Triple Points or Poor Wet Out

o Interfacial Separation

15
After Path Formation

16
CAF and Hollow Fibers
Hollow fibers form from decomposed impurities in the
glass melt

100 m

o Translation: It’s the glass

manufacturer’s fault

17
Hollow Fibers

o Generally, CAF is a two-step process

o Dependent on debonding between the glass fibers and epoxy resin
matrix to provide a path for copper migration

o With appearance of hollow fibers inside the laminates, CAF can happen
as a one step process.
o Concentration of hollow fibers in laminate becomes critical to reliability

18
Images of Hollow Fibers

Hollow fibers can be seen as white lines traveling along the

fiber bundle weave in both of these pictures.

19
Images of Hollow Fibers (cont.

20
How to Prevent Hollow Fibers

o Control your supply chain

o Laminate manufacturer can not change glass
supplier without prior approval

Who selects the glass fibers?

Laminate manufacturer?

Glass weave supplier?

21
How to Prevent Hollow Fibers (part 2)

o Qualify your supply chain

o Request the PCB manufacturer to periodically
perform hollow fiber assessment
o Hollow Fiber Assessment
o Burn-off the epoxy
o Cut the weave along the diagonal
(avoid double counting)
o Seal the edges with wax
o Immerse in an index-matching oil

o Goal is zero hollow fibers per 100 cm2

22
Drilling and CAF

o Drilling damage
can accelerate
CAF through
o Fiber/resin
delamination,
o Creation of
paths for
moisture to
accumulate
o Wicking due to
cracking of the
board material Drilling Damage

23
Wicking

Wicking can be serious if it extends sufficiently to deter the dielectric strength

or internal resistance breakdown between PTHs. It also provides a convenient
starting point for CAF as it effectively decreases the conductor spacing.

24
Avoiding Drilling Damage

o Drill bit manufacturers provide

PCB manufacturers guidance on
key process parameters
o Speeds and feeds
o Entry and exit material
o Number of drilling operations before repointing
o Stackup guidelines (number of PCBs of a given thickness that can
be stacked during drilling)
o Number of repoints / sharpening

o There is no ‘right’ answer for process parameters

o PCB manufacturer may buy a more expensive drill bit, but repoint
more often

25
Feeds / Speeds (example)

o Kyocera
o FR-4, Multilayer,
High Tg

o Chipload
o Feed / (Speed x
# of Flutes)

o SFM (surface feet

per minute)
o Feed x Dia. x 

26
Drilling (cont.)

o The key to success is verification/control and compliance/reward

o Did the PCB manufacturer perform their own DoE to understand and
verify guidelines from the drill bit manufacturer?
o Capture influence of high glass content, heavy copper, fill particles, etc.

o Are incoming and resharpened bits subjected to automatic

inspection?
o Are vacuum gauges alarmed and monitored?
o Is SPC of drill runout checked before production?

o Is the PCB manufacturer confirming employee compliance with

defined drilling parameters?
o Are bonuses inline with process parameters (compensation cannot be
increased by exceeding recommendations)

27
How Much Drilling Damage is Too Much?

o Current IPC specification is inadequate

o Acceptability of Printed Boards, IPC-A-600G, provides a target
and reject condition for haloing of unsupported holes only (not
plated through holes)

o IPC-A-600G does provide an accept / reject condition for

wicking
o Ranging from 80 to 125 microns, depending on class designation

o Wicking is the preferred defect assessment because

standard construction analysis is in a vertical orientation
o Haloing is easier to identify through non-standard horizontal
sectioning

28
Industry Specifications (cont.)

29
Debonding and Wetout

o Isola Group’s statement on laminate /

prepreg CAF performance
o “… wet out and interfacial chemistry
override other factors.”

30
 Poor Wet Out
Insufficient infiltration of epoxy
into glass weave can result in
‘triple points’

31
How to Prevent Wet-Out

o Glass spread

o Cleanliness of the glass fiber

before silane treatment

o Silane finish (coupling agent)

o Impregnation process parameters

(temperature, flow rates, etc.)

32
 Resin Fiber
Glass Style Glass
Style
Volume
Content
Volume
Content
1027 0.86 0.14

o PCB laminates (and prepregs) are 1037 0.86 0.14

106 0.84 0.16
fabricated with a variety of glass styles 1067 0.84 0.16
1035 0.83 0.17
1078 0.82 0.18
1080 0.79 0.21
1086 0.78 0.22
2313 0.74 0.26
2113 0.72 0.28
2116 0.71 0.29
3313 0.71 0.29

o The closed structure of low resin content 3070 0.68 0.32

1647 0.66 0.34
glass styles (e.g., 7628) can prevent 1651 0.66 0.34
adequate resin flow during impregnation 2165 0.66 0.34

o Can also trap ash after heat clean 2157 0.66 0.34
7628 0.64 0.36
(removal of starch-oil coating/sizing)

33
Glass Spread

o Trend is processes that improve wetting through spreading of

yarn bundles or opening capillaries between filaments.

o Top of the line PCB shops will require suppliers spec degree of
spreadness and provide a lot certification

34
Interfacial Separation
o Classic CAF is along the fiber/
epoxy interface

o Exposure to elevated temperature-

humidity conditions weakens glass
/polymer bonds based on silanes

o Hydrolysis reaction
o Si2O + H2O ↔ 2SiOH

o Attempts at improving the bonding

at this interface can result in
improved CAF performance

35
36
Silane Conundrum

o Classic Engineering Problem

o Properties good for one thing are not good for another
o The best choice is the most expensive

o Interface between the fiber and resin varies from tightly bound
siloxanes at the fiber wall to unbound siloxanes blending into
the epoxy matrix.
o Unbound siloxanes permit penetration of the epoxy resin into the
interface region and strengthens the epoxy-glass bond
o Tightly bound siloxanes restrict moisture absorption.

o The proper ratio of bound to unbound siloxanes results in the

optimum interface

37
Silane Conundrum (cont.)

o A method of increasing adhesion is to improve the

reactivity of surface treating agents

o Improved reactivity with resins can result in a rigid and

thin layer on the interfaces that can elevate residual stress

o The use of surface treating agents together with long chain

polysiloxanes will reduce the residual stress, but will tend
to decrease intrinsic interfacial adhesion

38
Silane Finale

o Dow Corning’s Z-6032 tends to dominate the market

o Vinylbenzylaminoethylaminopropyltrimethoxysilane
[C6H4-CH2-NHC2H4NHC3H6-Si(OCH3)3)]
o High water resistance and universal coupling agent
(pretty much works on all epoxy formulations)

o Problems
o Universal is not really universal (not all supply chains re-validate
compatibility with changes in resin)
o Z-6032 is expensive and instable (requires cooling); strong
motivation to select
lower cost options This is where your 5% price
reduction comes from!

39
Evidence of Non-Optimized Silane

Optical micrograph of copper filaments

in the area of fiber/resin delamination

Electron micrograph of area

of fiber/resin delamination.
EDS shows evidence of
copper filaments.
X-ray Map of Cu

40
Lot Qualification for Glass/Epoxy Bonding

o The process for ensuring good bonding at the glass/epoxy interface,

outside of CAF testing, is the use of the test method IPC TM-650 2.6.16
Pressure Vessel Method for Glass Epoxy Laminate.
o Requires exposing laminate coupons to pressure cooker conditions
(121C, 100%RH, 15 psi) for 30 minutes and then immersing the coupons
in a solder pot heated to either 260C or 288C.

o Called out in IPC-4101B Specifications for Base Materials for Rigid and
Multilayer Printed Boards
o Optional test
o If incorporated, IPC-4101B recommends that it be used for both
conformance and qualification testing, with testing performed on every
lot.

o The test method provides a grading system of 1 to 5, with laminates

graded 4 to 5 often rejected by the PCB supply chain

41
 What is Everyone Else
Doing About CAF?

42
Blissful Ignorance

o Most electronic OEMs do nothing in regards to CAF

o Blissful ignorance
o Not susceptible (design is too coarse to be an issue)

o Some electronic OEMs use ‘gut feel’

o Very conservative on voltage/spacing design rules
(graybeards)
o Require CAF-resistant laminate in PCB drawings
(could mean anything)
o Moved to more robust laminate with change to Pb-free
(higher Tg/Td/T288, lower moisture absorption)

43
Who is Concerned with CAF?

o Enterprise / Telecom / High-End Computing

o E.g., Cisco Systems, Sun (Oracle), IBM, etc.
o Aggressive Designs
o High I/O (>1000) BGAs with High Layer Count (18+) creates many
potential initiation sites
o Constant bias, but controlled environment
o Need for high availability

o Military / Avionics / Automotive / Industrial

o E.g., TRW, Rockwell Collins, etc.
o Modestly dense designs
o Uncontrolled environment
o Long life requirements (10+ years)

44
Preventing CAF

Design Supplier
Rules Qualification

45
CAF – Critical Paths

46
 Design Rules – Critical Paths

o Some debate over ‘critical’ paths

o PTH-to-PTH
o Potential for greatest internal
damage (2X drilling)
o Larger exposed surface area
o 20 to 30 mil spacings (12 mil drill
diameter minus 32 to 40 mil pitch)

o PTH-to-Plane
o Less potential for damage (1X drilling)
o 7 to 10 mil clearance

47
Proposed Trends in Conductor Pitch in PCBs

Determining the Limits of TelCordia Compliance for

Printed Wiring Boards by Karl Sauter
Circuitree, June 2000

Reduction in Pitch
100

Drilled Hole Size

80

Susceptibility of Glass-Reinforced Epoxy Laminates to

60
Conductive Anodic Filamentation by Chris Hunt

mils
Circuitree; March 1, 2007
40
When CAF?
Via
Edge 20
to
Via
Edge 0
1985 1990 1995 1999 2002 2004 2006

48
Current Industry Trends on Wall-to-Wall Spacing

o Conservative Designs
o Based on ball grid array (BGA) with 0.8mm (32mil) pitch
and a 0.3mm (12 mil) drill hole
o Wall-to-wall spacing of 0.5mm (20mil)

o Aggressive Designs
o Down to 0.25mm (10 mil) wall-to-wall spacing
o IPC Class 2 allows 100 microns of wicking
o Try to avoid spacings less than knuckle-to-knuckle distance
on the glass weave

49
 o A revolutionary automated design analysis tool that
brings insight and prediction earlier than ever into the
product development process

50
Sherlock and CAF Avoidance

o Since time-to-failure cannot realistically be determined for

CAF, Sherlock uses scoring to identify at risk designs
o 10 is in accordance with industry best practices
o 7 to 10 is designated green; indicates a preferred design
o 5 is in accordance with minimum acceptable practice
o 4 to 6 is designated yellow; indicates a marginal design
o 0 suggests a high likelihood of failure during lifetime
o 0 to 3 is designated red; indicates a high risk design

51
CAF Scoring

o Scoring is based on combination of wall-to-wall spacing,

degree of overlap (orthogonality), and qualification protocols
o Simple premise: The more aggressive the design, the more robust
the qualification method

o Examples
o Industry best practice (10) allows for 20mil spacing if each lot is
qualified
o Industry minimum practice (5) allows for 20 mil spacing if no
qualification is performed
o 10 mil spacing is a marginal, but not high risk, design (4) if each
lot is qualified
o Product qualification (design/material combination) is not
sufficient

52
Identifying At-Risk Sites for CAF

X1 (in) Y1 (in) Diam1 (mil) X2 (in) Y2 (in) Diam2 (mil) Distance (mil) Overlap (%)
13.02 4.365 12 13.02 4.39 12 13.0 100.0
11.745 3.565 12 11.745 3.59 12 13.0 100.0
14.61 4.5 12 14.635 4.5 12 13.0 100.0
14.61 4.53 12 14.635 4.53 12 13.0 100.0
9.65 4.62 12 9.675 4.62 12 13.0 100.0
10.245 2.58 12 10.27 2.58 12 13.0 100.0
13.025 4.25 12 13.05 4.25 12 13.0 100.0
13.11 5.025 12 13.135 5.025 12 13.0 100.0
8.97 2.735 12 8.98 2.76 12 14.9 16.7

53
CAF Scoring

54
Future Improvements

o Electric field
o Leveraged through existing use of net list
o Use CAF-resistant grade laminate
o Poorly defined at this time
o Comparison of laminate properties (Tg, Td, T288, moisture
absorption) to assembly temperatures
o Board thickness and stackup
o Presence of non-plated through-holes
o Parameters of CAF qualification test
o Supplier capability (‘sweet spot’)

55
Qualifying Suppliers: Select a Test Board

o IPC-9253 / IPC-9254 / PCQR2

o Wall-to-wall ranges from 10 to 25mil

o Alternative (should be based on your design rules!)

o Probably more common than IPC designs

o Some test boards qualify a specific design (zero failures)

o Some test boards assess margin (trying to cause failure)

o Material and stackup should be the same as actual product

56
Test Boards (cont.)

o IPC recommends 4200

initiation sites across
25 coupons

o Most company specs

require between 500 to
2000 initiation sites

57
CAF Test Coupon (Margining)

58
Qualifying Suppliers: Select a Test Condition
o Temperature / Humidity (sometimes with preconditioning)
o IPC TM-650, 2.6.25: 65C / 88%RH
o IPC-9151D (PCQR2): 75C / 85%RH
o IBM: 50C / 80%RH
o Others: 60C / 90%RH, 85C / 85%RH, etc.

o Voltage
o Not standardized
o Debate about high voltage (50V / 100V) vs. low voltage
(5V / 10V / 15V) and if bias voltage should equal test voltage
o IPC allows up to 100V (meets E-field limitation of 10V/mil)
o DfR Recommendation: Highest voltage (at appropriate spacing)
and smallest spacing (at appropriate voltage)

59
Qualifying Suppliers: Define Failure

o Driven by measurement approach

o Continuous monitoring (rare)
o Periodic (24 to 72 hours)

o Failure definition
o Resistance (100 megaohms)
o Change in resistance (10X)

o Define test time

o Enterprise: 300 to 600 hours
o Automotive: 500 to 2000 hours

60
Defining Time to Failure (IPC-9691)

61
CAF is an Infant Mortality Defect
Test Condition (85C / 85%RH)

62
CAF in the Future

o Will the risk of CAF increase in future designs?

o Not really

o The rate of feature size reduction is very limited with PCBs

o Industry roadmaps have barely moved (except for
substrates, which tend to rely on laser drilling and resin-
coated copper without glass fibers)

o Pb-free transition is almost complete

o Some movement to lower temperature alloys (e.g., SnBi)

63
 PCB Industry Plated Through Hole Capability

95% of Industry
95% of Industry

5% of Industry
5% of Industry

1.2 – 1.6mm PCB Thickness 2.3 – 3.2mm PCB Thickness

IBM PCB-OS Symposium 2007, Roadmap Technology Verification,

DDI Tech Conductor Analysis Technologies (CAT)
PCB Thickness (mil / mm)
Roadmap 2011
Via Diameter
(mil / um) Standard Advanced Engineering
o Minimal technological
6 / 150 N/A 39 / 1.0 60 / 1.5
progress over past 10
8 / 200 64 / 1.6 80 / 2 96 / 2.4
years
10 / 250 100 / 2.5 120 / 3 160 / 4

64
Summary / Conclusion

o Conductive Anodic Filament (CAF) formation does happen

o When it happens, it can cause a lot of pain

o CAF behavior is relatively stable

o Limited change in key PCB technology (pitch, materials,
assembly)

o CAF mitigation is well known (execute it!)

o Evaluate your designs
o Qualify your suppliers

65
Disclaimer & Confidentiality
o ANALYSIS INFORMATION
This report may include results obtained through analysis performed by DfR Solutions’ Sherlock
software. This comprehensive tool is capable of identifying design flaws and predicting
product performance. For more information, please contact sales@dfrsolutions.com.

o DISCLAIMER
DfR represents that a reasonable effort has been made to ensure the accuracy and reliability
of the information within this report. However, DfR Solutions makes no warranty, both express
and implied, concerning the content of this report, including, but not limited to the existence of
any latent or patent defects, merchantability, and/or fitness for a particular use. DfR will not
be liable for loss of use, revenue, profit, or any special, incidental, or consequential damages
arising out of, connected with, or resulting from, the information presented within this report.

o CONFIDENTIALITY
The information contained in this document is considered to be proprietary to DfR Solutions
and the appropriate recipient. Dissemination of this information, in whole or in part, without
the prior written authorization of DfR Solutions, is strictly prohibited.

From all of us at DfR Solutions, we would like to thank you for choosing us as your partner in
quality and reliability assurance. We encourage you to visit our website for information on a wide
variety of topics. To help us continually improve, please send any feedback or comments to
iso@dfrsolutions.com.
Best Regards,
Dr. Craig Hillman, CEO

66