Selecting the Right Mitigation for BGAs and QFNsresources.dfrsolutions.com/Webcasts/2016/Mitigation-of-QFN-and-BGA... · BGAs o However, far less substantive research on the influence

© 2004 - 2007© 2004 - 20109000 Virginia Manor Rd Ste 290, Beltsville MD 20705 | 301-474-0607 | www.dfrsolutions.com© 2004 – 2010

Selecting the Right Mitigation for BGAs and QFNs

DfR Webinar

March 31, 2016

© 2004 - 2007© 2004 - 20109000 Virginia Manor Rd Ste 290, Beltsville MD 20705 | 301-474-0607 | www.dfrsolutions.com

o Ball grid array (BGA*) and quad flat pack no-lead (QFN)

are increasingly the packaging of choice for most

integrated circuit (IC) devices

Motivation

*Includes chip scale packaging (CSP)


BGAs: Almost 2/3rd of Semiconductor Packaging Revenue


QFNs: Most Popular IC Packaging by Volume


o Thermal Performance

o qja for the QFN is half of leaded counterparts (2X increase in power dissipation)

o Electrical Performance

o Inductance for QFN ishalf of leaded counterparts (no leads, shorter wire bonds)

o Size and Cost

o ‘Poor Man’s BGA’

Why QFNs?


Why BGAs? The Right Packaging can be Worth $2.5 Billion

www.valuewalk.com/2015/12/apple-iphone-7-a10-chip-solely-tsmc/

© 2004 - 2007© 2004 - 20109000 Virginia Manor Rd Ste 290, Beltsville MD 20705 | 301-474-0607 | www.dfrsolutions.com7

Why Not BGAs and QFNs?

o Greater risk of solder joint failure (thermal cycling,

vibration, mechanical shock)

Cycles to failure

-40 to 125C QFP: >10,000

BGA: 3,000 to 8,000

QFN: 1,000 to 3,000CSP / Flip Chip: <1,000


o Option 1: Don’t Use Them

o Increasingly not an option

o Option 2: Physics of Failure

o Change the design; predict time to failure

and change the placement, PCB materials,

or housing to meet reliability goals

o Option 3: Underfill/Staking

o Don’t change the design; apply a

polymeric material to reduce strain

on the solder joint

How to Mitigate Risk of QFNs and BGAs


Background


o The U.S. Navy and multiple companies funded DfR

Solutions to develop deep expertise in the performance of

Pb-free solders under harsh environments

o Thermal Cycling, Vibration, Mechanical Shock

o Through two SBIRs (Phase I and II)

o Goals

o Develop a methodology to predict vibration and mechanical shock

performance over a range of temperatures (not just 25°C)

o Identify optimum mitigation strategies (corner stake, edge bond,

underfill)

Background


o Four (4) loading conditions of concern to modern

electronics

o Thermal Cycling (low-cycle fatigue)

o Mechanical Shock / Drop

o Bending (Cyclic or Overstress)

o Vibration (high-cycle fatigue)

Background


Thermal Cycling Failure Behavior

o Driven by different expansion/contraction behaviors

o Because solder is connecting two materials that are expanding /

contracting at different rates (GLOBAL)

o Because solder is expanding / contracting at a different rate than

the material to which it is connected (LOCAL)


o This differential expansion and contraction introduces

stress into the solder joint

o This stress causes the solder to deform (aka, elastic and plastic

strain)

o The extent of this strain

(that is, strain range or

strain energy) tells us the

lifetime of the solder joint

o The higher the strain, the

more the solder joint is damaged,

the shorter the lifetime

Thermal Cycling Failure Behavior (cont.)


o Low cycle fatigue (LCF) is driven by inelastic strain (Coffin-Manson)

o Difficult to provide predictions under 100 events/cycles

o Considered relevant out to 10,000 cycles

o High cycle fatigue (HCF) is driven by elastic strain (Basquin)

o Primarily vibration, but can be bending as well

o Failures above 100,000 cycles

Mechanical Cycling (Shock, Bending, Vibration)

b

f

f

e NE

2

-0.05 < b < -0.12; 8 < -1/b < 20


Option 2: Physics of Failure


o Knowing the critical drivers for solder joint fatigue, we can

develop predictive models and design rules

Drivers for Solder Joint Thermo-Mechanical Failures

CTE of Board

Elastic Modulus (Compliance) of Board

CTE of Component

Elastic Modulus (Compliance) of Component

Length of Component

Volume of Solder

Thickness of Solder

Solder Fatigue Properties


Predictive Models – Physics of Failure (PoF)

o Modified Engelmaier for Pb-free Solder (SAC305)

o Semi-empirical analytical approach

o Energy based fatigue

o Determine the strain range (Dg)

o C is a correction factor that is a function of dwell time and

temperature, LD is diagonal distance, a is coefficient of

thermal expansion (CTE), DT is temperature cycle, h is

solder joint height

Th

LC

s

D DDD ag


Predictive Models – Physics of Failure (PoF)(cont.)

o Determine the shear force applied to the solder joint

o F is shear force, L is length, E is elastic modulus, A is the area, h

is thickness, G is shear modulus, and a is edge length of bond pad

o Subscripts: 1 is component, 2 is board, s is solder joint, c is bond

pad, and b is board

o Takes into consideration foundation stiffness and both

shear and axial loads

D

aGGA

h

GA

h

AE

L

AE

LFLT

bcc

c

ss

s

9

2

221112

aa


Predictive Models – Physics of Failure (PoF)(cont.)

o Determine the strain energy dissipated by the

solder joint

o Calculate cycles-to-failure (N50), using energy

based fatigue models

10019.0

D WN f

sA

FW DD g5.0

© 2004 - 2007© 2004 - 20109000 Virginia Manor Rd Ste 290, Beltsville MD 20705 | 301-474-0607 | www.dfrsolutions.com20

Validation of Physics of Failure (Thermal Cycling)

BGA

QFN


o “The Sherlock analysis that DfR solutions did last year predicted

that the QFN would fail first during thermal cycling <sic>. At the

time last year we did not agree with what Sherlock was telling us,

but after lots of $$$ and thermal cycles, Sherlock was right.

It was a very important lesson learned about thermal cycling life

of large QFN packages on boards that are epoxy laminated to

aluminum chassis<sic>. This is something that we would not have

guessed to be a problem. But it was.

Sherlock now has added merit for us to consider using on other

new product reliability analyses”

- John DeCamp, Manufacturing Engineer

Validation – ViaSat

21


o “Following discussions with DfR and confirmation of the PCB material, from the supplier, I was able to refine the model for the QFN package and the PCB construction to predict the first failure of a QFN package at around 700 cycles.

It was interesting to note that the PCB material choice significantly altered Sherlock’s predicted solder joint life and choice of PCB material needs to be carefully considered from this perspective.

Subsequent real cycling of the test board has indeed produced a failure at around 770 temperature cycles and so appearing to add some (albeit limited) validation of the Sherlock prediction.

With the refined model failures of well soldered BGA joints were not predicted by Sherlock till around 3000 cycles. Our supplier has now finished the thermal cycles on the real boards seeing no BGA failures after about 1200 cycles.”

Validation – Rolls Royce

22


Influence of PCB Properties


o During vibration, board-level strain is

proportional to solder or lead strains

and therefore can be used to make

time-to-failure predictions

o Requires converting

cycles-to-failure displacement

equations (Steinberg) to use strain

o The critical strain for the package

types is a function of package style,

size, lead geometry

High Cycle Fatigue (Prediction)

n

ccNN

0

0

Lcc

ζ is analogous to 0.00022B but

modified for strain

c is a component packaging function

L is component length


o Incorporate a 2x safety factor per Steinberg (f reduced by 50%)

Validation of Physics of Failure (Vibration)

Cycles to

Failure

SAC305SnPbSolder Alloy

4.6

116

fN

8.7

71

fN


o Step 1: Implement in Component Engineering

o Each new request to add a part to the approved vendors list (AVL) should be assessed for risk of solder joint failure (ie, all QFNs and BGAs)

o Step 2: Benchmark PoF Prediction

o Run a PoF simulation and compare to supplier’s test data

o If no test data, identify reliability by similarity (RBS) equivalent

o If RBS, decide to either test or move forward with PoF only

o Step 3: Perform PoF before Layout

o After layout is too late to change the design (stackup, bond pad sizing, stencil thickness, placement, attachments, etc.)

Developing a PoF Process


Option 3: Underfill/Staking

(Thermal Cycle Mitigation)


o Underfill has been used since to mitigate flip chip solder

risk since the late 1980’s

o Up to two orders of magnitude improvement in lifetime

o Has led to interest in mitigating thermal cycling risks with

BGAs

o However, far less substantive research on the influence of underfill

and thermo-mechanical fatigue of 2nd level interconnects

o General trends can be identified through review of

published work and the use of physics of failure (PoF)

Thermal Cycle Mitigation: BGA


o BGA

o 12.5MM X 12.5MM X 1.2MM

o 900 Solder Balls (0.4mm pitch, 0.25mm dia., 0.11mm height)

o Die size 8MM X 8MM

o Thermal Cycle

o 15C to 85 C, 10 Hours cycle

o PCB

o CTE of 15.8 ppm/°C

o Modulus of 44 GPa

Step 1: PoF-Based Simulation on BGA with No Underfill

Sherlock predicts a

lifetime of 1500 cycles


o Glass Transition Temperature (Tg) of 102°C

o Coefficient of Thermal Expansion (CTE) of 55 ppm/°C

o Elastic Modulus (E) is 3 GPa

Step 2: Adjust PoF Prediction Based on Underfill Properties


o Tg has limited influence as long as it is greater than

maximum temperature during thermal cycling

o Derived approach

o Assumed adjustment factor

is 1 for Tg above 100°C

and 0.1 for Tg at 25C

o Assume a logarithmic

response to change in Tg

o Based on thermal cycle

of 0 to 100C

Tg and Life Adjustment Factor

y = 0.6ln(x) - 1.8

0

0.2

0.4

0.6

0.8

1

1.2

0 20 40 60 80 100 120 140

Fati

gue

Life

Co

rrec

tio

n F

acto

r

Tg in Degrees Celsius

Tg Fatigue Correction Factor


o Underfill CTE is almost always greater than 20 ppm/°C

CTE and Life Adjustment Factor

Figure 8 From “A Parametric Study of Flip Chip Reliability Based on Solder

Fatigue Modelling: Part II - Flip Chip on Organic” by Scott F. Popelar

y = 9000x-2

1

10

100

1000

10000

0 10 20 30 40 50 60

Fati

gue

Lif

e C

orr

ect

ion

Fac

tor

Underfill CTE (ppm/C)

CTE Fatigue Life Correction

Model doesn’t work for values below CTE of solder


o Since most underfill have modulus close or near 2 GPa

the fatigue life correction factor at 2 is set to near 1

Modulus and Life Adjustment Factor

Figure 7 From “A Parametric Study of Flip Chip Reliability Based on Solder

Fatigue Modelling: Part II - Flip Chip on Organic” by Scott F. Popelar

Model doesn’t work for values below 2

y = e-0.05x

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 5 10 15 20

Fatigue

Lif

e C

orr

ect

ion

Fact

or

Modulus (GPa)

Modulus Fatigue Life Correction


o Maximum temperature of 85°C below Tg of 102C

o CTE is relatively high 55 ppm/ C

o Modulus is nominal 3 GPa

o Prediction formula (not PoF-based)

𝑒−0.05𝐸 × 9000g−2 × 0.6 ln 𝑇𝑔 − 1.8Where

E = modulus in GPa

g = CTE in ppm/ C

Tg = Glass Transition Temperature in C

Life Adjustment Factor


o Predicted life increase using underfill is 2.5X

o Sherlock Prediction (no underfill): 1500 cycles

o Prediction with Adjustment Factor (underfill): 3750 cycles

Life Adjustment Factor: Results


o Underfill is not always the best optiono Expensive

o Time-consuming

o Low throughput

o Increasing interest in the use of edge bonding to provide a similar degree of protectiono Or, at the very least, not to reduce time to

failure (if edge bond is used for vibration or shock mitigation)

o Evaluated through thermal cycle testingo -55C to +125C (thermal cycle), 15 min dwells,

1000 cycles

o Mid-size BGAs (256 and 288 I/O)

Thermal Cycle Mitigation: Edge Bonding


o Minimal to no improvement [66 ppm/C, Two Tgs (-39 and 77)]

0%

20%

40%

60%

80%

100%

120%

0 200 400 600 800 1000

thermal cycles

cu

mu

lati

ve f

ail

ure

Choc. Chip, Scotchweld 2214

Edge Bond, EMAX 904-GEL

Edge Bond, Loctite 3128


Edge Bond, Namics 1583-2

Edge Bond, Scotchweld 2214

Edge Bond, Zymet UA-2605

As Manufactured

A-CABGA56-.5mm-6mm-DC


A-CABGA256-1.0mm-17mm-DC

no underfilled parts failed

B Edge Bond (not reworkable)


C Edge Bond

o Minimal to no improvement [40 ppm/C, Tg of 45C]

0%

20%

40%

60%

80%

100%

120%

0 200 400 600 800 1000

thermal cycles

cu

mu

lati

ve f

ail

ure








As Manufactured






D Edge Bond

o 50% improvement [23 ppm/C, Tg of 80C]

0%

20%

40%

60%

80%

100%

120%

0 200 400 600 800 1000

thermal cycles

cu

mu

lati

ve f

ail

ure








As Manufactured






0%

20%

40%

60%

80%

100%

120%

0 200 400 600 800 1000

thermal cycles

cu

mu

lati

ve f

ail

ure








As Manufactured





A Edge Bond (Reworkable)

o Substantial improvement [30 ppm/C, Tg of 90C]


o The lower the CTE and the higher the Tg, the better the

mitigation

o Smaller BGAs may be at some risk of reduced time to

failure

o 6mm BGA (not shown in graphs) failed slightly faster with edge

bonding

o Staking can potential provide the necessary mitigation at

substantially lower cost and higher throughput

Thermal Cycle Migitation (Underfill/Staking): Conclusion


Option 2: Underfill/Staking

(Vibration/Shock Mitigation)


o SBIR:

Assess vibration and shock performance over a range of

temperatures

o SBIR Consortium:

Identify optimum mitigation strategies for vibration and

mechanical shock (corner stake, edge bond, underfill)

Research Focus


SBIR: Experimental Setup, Parts

Package Style Package Type Pitch Manufacturer

Chip Component 1206 0W resistor N/A Panasonic

Chip Component 2512 0W resistor N/A Stackpole

Chip Component 1812 120W ferrite bead N/A TDK

C-Lead SMC 3126 C-Lead Schottky Diode N/A Fairchild

CSP (chip scale package) CSP97 (CVBGA97) 0.4 Practical

Mid-Size BGA (ball grid array) BGA484 (PBGA484) 1.0 Practical

Large BGA (ball grid array) BGA1156 (PBGA1156) 1.0 Practical

TSOP Type I (thin small-outline package) TSOP48, Type I 0.5 Topline Electronics

TSOP Type II (thin small-outline package) TSOP54, Type II 0.8 Topline Electronics

QFP (quad flat pack) QFP100 0.5 Practical

QFN (quad-flat no-leads) QFN68 0.5 Practical


o Mechanical Shock at 125°C

was implemented through the

use of in-board heating traces

o Mechanical Shock at -55°C

was implemented through

the use of an insulated

box attached to the drop

plate

SBIR: Experimental Setup, Environments (cont.)


o Because extensive sinusoidal vibration testing was

performed on a range of components in SBIR Phase I,

Sinusoidal Vibration was limited to one displacement

condition

o Sinusoidal Vibration model developed and validated

o Executed on a modified Unholtz-Dickie

SBIR: Experimental Setup, Environments (cont.)

Condition Displacment Frequency* Temperature (°C)

1 90 mils 160-165 Hz 25

2 90 mils 160-165 Hz 125

3 90 mils 160-165 Hz (-55)

*5 to 10Hz below natural frequency


o Frequency range was adjusted to account for board response and the tendency of the component to fail under vibration

o PSD curve was applied for 80 Hz above and 80 Hz below the natural frequency of the board


BGA484 and BGA1156: Experimental Design

Test Temp Stress Level Finish Solder Alloys

Shock 25°C 550G ENIG SnPb, SAC305, SAC105

Shock 25°C 550G OSP SnPb, SAC305, SAC105





Random 25°C 0.1G2/Hz OSP SAC105

Random 125°C 0.1G2/Hz OSP SnPb, SAC305, SAC105

Random (-55)°C 0.1G2/Hz OSP SnPb, SAC305, SAC105

Random 25-125°C 0.1G2/Hz OSP SnPb, SAC305, SAC105

Test Temp Stress Level Finish Solder Alloys









Shock (-55)°C 2500G ENIG SnPb, SAC305, SAC105

Shock (-55)°C 2500G OSP SnPb, SAC305, SAC105

Sinusoidal 25°C 90mils OSP SAC105

Sinusoidal 125°C 90mils OSP SAC105

Sinusoidal (-55)°C 90mils OSP SAC105



Random (-55)°C 0.1G2/Hz OSP SnPb, SAC305, SAC105

Crack Initiation (-55)°C – 125°C ENIG SnPb, SAC305, SAC105

Crack Initiation (-55)°C – 125°C OSP SnPb, SAC305, SAC105

Crack Initiation 25°C – 125°C ENIG SnPb, SAC305, SAC105

Crack Initiation 25°C – 125°C OSP SnPb, SAC305, SAC105

BGA484 BGA1156


o All failures were solder joint failures

o Identified very clear trends in regards to solder alloy and

temperature

BGA484 and BGA1156: Mechanical Shock


o SAC105 and SAC305 show strong sensitivity to temperature

o Minimal difference between OSP / ENIG


o SnPb solder clearly superior, even at room temperature,

to SAC105 and SAC305 at high strain rates

o Minimal difference between OSP and ENIG

BGA 1156: Mechanical Shock


o A temperature dependent failure trend can be seen, with hotter temperatures failing first. o At higher temperatures the crack propagates through the bulk solder, while

at lower temperatures it propagates along the IMC.

BGA484: Sinusoidal Vibration


o Able to create design curve from test results from SBIR

Phase I and mechanical shock and vibration test results

BGA484 and BGA1156: Design Curves


o Components ‘generally’ behaved as expected

o Either already known or explained based on mechanical response

or material behavior

o Allows for the development of a vibration as a function of

temperature capability in Sherlock

o Lower stresses tends to reduce differentiation among

solder alloys

o True for both mechanical shock and vibration

Overall Trends


o Edge bond/Corner Stake: Zymet & Namics

o Underfill: Loctite & Namics

Experimental Design – Mitigation (BGA208)

Corner Stake Edge Bond Underfill


o Mechanical Shock:

o 1500G, 0.5 millisecond pulse

o -55C, 25C, 125C

o Sinusoidal Vibration

o 90 mils displacement

o Situated 5-10 Hz below natural frequency

Experimental Design


Shock/Drop and Corner StakingReliaSoft Weibull++ 7 - www.ReliaSoft.com

Probability - Weibull

Corner Stake BGA\1583: Corner Stake BGA\UA 2605: Corner Stake BGA\SnPb: Corner Stake BGA\SN100C: Corner Stake BGA\SAC:

Number of Drops

Un

reli

ab

ilit

y,

F(

t)

10.000 1000.000100.0001.000

5.000

10.000

50.000

90.000

99.000

x 4

Probability-Weibull

Corner Stake BGA\SACWeibull-2PRRX SRM MED FMF=12/S=3

Data PointsProbability L ine

Corner Stake BGA\SN100CWeibull-2PRRX SRM MED FMF=10/S=5


Corner Stake BGA\SnPbWeibull-2PRRX SRM MED FMF=10/S=5


Corner Stake BGA\UA 2605Weibull-2PRRX SRM MED FMF=4/S=11


Corner Stake BGA\1583Weibull-2PRRX SRM MED FMF=6/S=9


Melissa KeenerDfR Solutions8/7/20123:55:11 PM


Shock/Drop and Edge BondingReliaSoft Weibull++ 7 - www.ReliaSoft.com


Edge Bond BGA\1583: Edge Bond BGA\VA2605: Edge Bond BGA\SN100C: Edge Bond BGA\SnPb: Edge Bond BGA\SAC:

Number of Drops

Un

reli

ab

ilit

y,

F(

t)

10.000 1000.000100.0001.000

5.000

10.000

50.000

90.000

99.000

x 4

Probability-Weibull

Edge Bond BGA\SACWeibull-2PRRX SRM MED FMF=12/S=3


Edge Bond BGA\SnPbWeibull-2PRRX SRM MED FMF=10/S=5


Edge Bond BGA\SN100CWeibull-2PRRX SRM MED FMF=10/S=5


Edge Bond BGA\VA2605Weibull-2PRRX SRM MED FMF=7/S=8


Edge Bond BGA\1583Weibull-2PRRX SRM MED FMF=6/S=9


Melissa KeenerDfR Solutions8/14/20123:57:00 PM


Shock/Drop and UnderfillReliaSoft Weibull++ 7 - www.ReliaSoft.com


Underfill\BGA L3549: Underfill\SN100C: Underfill\SnPb: Underfill\SAC305:

Number of Drops

Un

reli

ab

ilit

y,

F(

t)

10.000 1000.000100.0001.000

5.000

10.000

50.000

90.000

99.000

x 4

x 15

Probability-Weibull

Underfill\SAC305Weibull-2PRRX SRM MED FMF=12/S=3


Underfill\SnPbWeibull-2PRRX SRM MED FMF=10/S=5


Underfill\SN100CWeibull-2PRRX SRM MED FMF=10/S=5


Underfill\BGA L3549Weibull-1PMLE SRM MED FMF=0/S=15

Susp PointsProbability L ine

Melissa KeenerDfR Solutions8/31/201211:20:58 AM


Mitigation Technique Mitigation Material Benefit

Corner StakingA 3 times better than baseline

C 2 times better than baseline

Edge BondingA 10 times better than baseline


UnderfillB No failures during testing

D 10 times better than baseline

Effect of Mitigation: Mechanical Shock


Corner StakingA 100 times better than baseline

C No change from baseline



UnderfillB 10 times better than baseline

D >100 times better than baseline


Corner StakingA No change from baseline






-55C

25C

125C



Corner StakingA No change from baseline




UnderfillB No failures during testing


Effect of Mitigation: Sinusoidal Vibration


Corner StakingA No failures during test


Edge BondingA No failures during test

C No failures during test


D No failures during test


Corner StakingA 20 times worse than baseline

C 20 times worse than baseline

Edge BondingA No change from baseline

C 40 times worse than baseline

UnderfillB No change from baseline

D One failure during testing

-55C

25C

125C


o The value of mitigation can be quantified

o Sherlock now offers coating/potting/staking module

o Simple design solutions (underfill everything) can be

evaluated against simple manufacturing solutions (change

stackup, bond pads)

o Sometimes less is more

o Sometimes less is not enough!

Conclusion

Coating/Potting

Documents

Selecting the Right Mitigation for BGAs and QFNsresources.dfrsolutions.com/Webcasts/2016/Mitigation-of-QFN-and-BGA... · BGAs o However, far less substantive research on the influence