IPC-D-279 EN.pdf - 第49页

pad centers, L D is sometimes referred to as the distance from the neutral point (DNP); T C ,T S = steady-state operating temperature for compo- nent, substrate (T C >T S for power dissipation in component) during hig…

Solder, uniquely among the commonly used engineering

metals, readily creeps and stress-relaxes at normal use tem-

peratures; the rate of creep and stress-relaxation is highly

temperature- and stress-level-dependent. Thus, the cyclic

fatigue damage term, ∆D, for practical reasons, has to be

based on the total potential damage at complete creep/

stress relaxation of the solder. For cyclic conditions that do

not allow sufficient time for complete stress relaxation ∆D

is larger than the actual fatigue damage. The temperature-

and time-dependent exponent, c, compensates for the

incomplete stress relaxation and is given by

c =−0.442 − 6x10

−4

+ 1.74x10

−2

ln(1 +

360

)

[Eq. A-2]

where

= mean cyclic solder joint temperature

= half-cycle dwell time in minutes.

The half-cycle dwell time relates to the cyclic frequency

and the shape of the cycles and represents the time avail-

able for the stress-relaxation/creep to take place.

In Eq. A-1 the exponent is given as (−1/c), which is men-

tally confusing; this format exists for historical reasons in

that the underlying work [Refs. A-9: 14,15] was always

stated this way. For typical electronic applications (T

to 100°C and t

= 15 to 720 minutes) the exponent (−1/c

ranges between 2.0 and 2.6.

Equations A-1 and A-2 come from a generic understanding

of the response of SM solder joints to cyclically accumu-

lating fatigue damage resulting from shear displacements

due to the global thermal expansion mismatches between

component and substrate. These shear displacements cause

time-independent yielding strains and time-, temperature-,

and stress-dependent creep/stress relaxation strains. These

strains, on a cyclic basis, form a visco-plastic strain energy

hysteresis loop which characterizes the solder joint

response to thermal cycling and whose area, given as the

damage term ∆D, is indicative of the cyclically accumulat-

ing fatigue damage. Hysteresis loops in the shear stress-

strain plane have been experimentally obtained [Refs. A-9:

13,17-19].

A-3.2 Damage Modeling

The assessment of the cyclically cumulating fatigue dam-

age is not a straight-forward task. While Eq. A-1 is widely

used, the question of how to best quantify the cyclic fatigue

damage is still hotly debated. The choices are primarily

between more complex ﬁnite-element analyses (FEA),

which can give more detailed information and can include

second-order effects, but require a large number of not

fully-supported assumptions [Ref. A-9: 20]; and closed-

form empirically-based relationships of the ﬁrst-order

design parameters, which cannot include second-order

effects and have use limitations due to their simple nature,

but allow, due to their simple form, a direct assessment of

the impact of the primary design parameters as well as

design trade-offs.

The following cyclic fatigue damage terms are of the sim-

pliﬁed closed-form type and should be utilized with the

application caveats that follow [Refs. A-9: 1-6,9,12,16,21].

The cyclic fatigue damage term for leadless SM solder

attachments, for which the stresses in the solder joints

exceed the solder yield strength and cause plastic yielding

of the solder, is

∆D(leadless)=

[

∆(α∆T)

]

[Eq. A-3]

For solder attachments with leads compliant enough, so

that the solder joint stresses are below the yield strength

and thus are not bounded by it, the cyclic fatigue damage

term is

∆D(leaded)=

[

∆(α ∆T)]

(919 kPa)Ah

]

[Eq. A-4]

where for English units the scaling coefficient is 133 psi

instead of 919 kPa.

Equation A-4 contains the design parameters that have a

ﬁrst-order inﬂuence on the reliability of SM solder attach-

ments. They are

A = effective minimum load bearing solder joint

area;

F = empirical ‘‘non-ideal’’ factor indicative of

deviations of real solder joints from idealizing

assumptions and accounting for secondary and

frequently intractable effects such as cyclic

warpage, cyclic transients, non-ideal solder

joint geometry, different solder crack propaga-

tion distances, brittle IMCs, Pb-rich boundary

layers, and solder/bonded-material expansion

differences, as well as inaccuracies and uncer-

tainties in the parameters in Eqs. A-1 through

A-4; 1.5>F> l.0 for ball/column-like leadless

solder joints (C4, C5, BGAs, CGAs),

1.2>F>0.7 for leadless solder joints with fillets

(castellated chip carriers and chip compo-

nents), F≈1 for solder attachments utilizing

compliant leads;

h = solder joint height;

= ‘‘diagonal’’ ﬂexural stiffness of unconstrained,

not soldered, corner-most component lead,

determined by strain methods [see Refs. A-9:

22-25] or FEA;

= maximum distance between component solder

joints measured from component solder joint

July 1996 IPC-D-279

pad centers, L

is sometimes referred to as the

distance from the neutral point (DNP);

= steady-state operating temperature for compo-

nent, substrate (T

for power dissipation in

component) during high temperature dwell;

C,0

S,O

= steady-state operating temperature for compo-

nent, substrate during low temperature dwell,

for non-operational (power off) half-cycles

C,O

S,0

;

= (1/4)(T

C,O

S,O

), mean cyclic solder

joint temperature;

= CTEs for component, substrate;

∆D = potential cyclic fatigue damage at complete

stress relaxation;

∆T

−T

C,O

, cyclic temperature swing for compo-

nent;

∆T

−T

S,O

, cycling temperature swing for sub-

strate (at component);

∆(α∆T) = ?α

∆T

−α

∆T

?, absolute cyclic expansion

mismatch, accounting for the effects of power

dissipation within the component as well as

temperature variations external to the compo-

nent;

∆α = ?α

−α

?, absolute difference in CTEs of com-

ponent and substrate, CTE-mismatch, because

of the inherent variability in material proper-

ties ∆α<2x10

-6

should not be used in calculat-

ing reliability.

A-3.3 CAVEAT 1 — Solder Joint Quality

The solder joint fatigue behavior and the resulting reliabil-

ity prediction equations. Eqs. A-1 through A-2, were deter-

mined from thermal cycling results of solder joints that

failed as a result of fracture of the solder, albeit sometimes

close to the IMC layers. For solder joints for which layered

structures are interposed between the base material and the

solder joints, these equations could be optimistic upper

bounds if the interposed layered structures become the

‘weakest link’ in the surface mount solder attachments.

Such layered structures could be: metallization layers that

have weak bonds to the underlying base material, or are

weak themselves, or dissolve essentially completely in the

solder; oxide or contamination layers preventing a proper

metallurgical bond of the solder to the underlying metal;

brittle IMC layers too thick due to too many or improperly

long high temperature processing steps.

Some material choices can lead to lower quality and

weaker solder joints because the material is more difficult

to wet and solder. The nickel/iron alloys, Kovar

and

Alloy 42, fall into this material category. The resulting

lower solder joint quality indicated in Table A-2 is also

clearly evident in Figure A-2, where the solder joint pull

strength is shown for a variety of differently prepared

Alloy 42 and copper leads. Alloy 42 leads, even when

etched or pre-ﬂowed at temperatures higher than can be

tolerated by the component, show a substantial reduction in

the solder joint pull strength relative to copper. In the worst

instance, the leads from one Alloy 42 manufacturer have a

pull strength of less than half of those with more typical

Alloy 42 and are essentially non-wettable.

Early failures of the solder attachments of components with

Alloy 42 lead frames and leads during accelerated testing

[Refs. A-9: 26-30] and manufacture [Ref. A-9: 31] have

been documented.

Solder joints which have solder joint heights (gaps) of

h<50 to 75µm also require special attention. For solder

joints that thin, the gap is essentially ﬁlled with intermetal-

lic compounds and those solder metals that do not go into

solution with the base metals to form the IMCs. Therefore

Eqs. A-1 and A-2 do not apply because these gaps are not

ﬁlled with solder [Ref. A-9: 32]. These materials do not

creep as readily, if at all, at the prevailing temperatures and

are typically more brittle, but much stronger than solder.

Thus, fatigue lives are longer than would be predicted from

Eqs. A-1 and A-2 unless overstress conditions occur.

On the other hand, the fatigue lives of solder attachments

can be underestimated by Eqs. A-1 through A-4 if the com-

ponent is underﬁlled with a load-bearing substance, e.g.,

epoxy [Ref. A-9: 33]. Components that are glued-down to

the substrate result in higher solder joint fatigue reliability,

since the solder joints are loaded in compression when the

adhesive contracts on cooling from the solder reﬂow tem-

peratures. Covercoats can either increase or decrease solder

joint fatigue lives depending on the properties of the cover-

coat and when and how it is applied. Parylene

has been

found to increase the solder joint fatigue life by about a

factor of three.

In general, caution might be indicated in all instances

where the predicted life is less than 1000 cycles, because

the severe loading conditions producing such short lives

are likely to produce different damage mechanisms or/and

failure modes.

A-3.4 CAVEAT 2 — Large Temperature Excursions

Solder joints experiencing large temperature swings

(−50°C to + 80°C) which extend both signiﬁcantly below

and signiﬁcantly above the temperature region bounded by

Table A−2 Quality of Solder Joints with Copper and

Alloy 42 Resulting from Different Reﬂow Temperatures

Reﬂow

Temperature

°C

Solder Joint Quality

60/40 Solder to

Copper

60/40 Solder to

Alloy 42

~210 just o.k. marginal to bad

~240 good just o.k.

~260 good good

IPC-D-279 July 1996

−20°C to +20°C, in which the change from stress- to

strain-driven solder response takes place, do not follow the

damage mechanism described in Eqs. #1 and #2 [Ref. A-9:

34]. The damage mechanism is different than for more

typical use conditions and is likely dependent on a combi-

nation of creep-fatigue, causing early micro-crack forma-

tion, and stress concentrations at these micro-cracks caus-

ing faster crack propagation during the high stress cold

temperature excursions, as well as recrystallisation consid-

erations.

A-3.5 CAVEAT 3 — High-Frequency/Low-Temperatures

For high-frequency applications, f>0.5 Hz or t

<l s, e.g.,

vibration, and/or low temperature applications, T

< 0°C,

for which the stress relaxation and creep in the solder joint

is not the dominant mechanism, the direct application of

the Coffin-Manson [Ref. A-9: 14] fatigue relationship

might be more appropriate. This relationship is

(50%)=

[

f’

∆γ

]

−1

[Eq. A-5]

where ∆γ

is the cyclic plastic strain range and c ≈ −0.6.

It has to be noted, that the determination of ∆γ

depends on

the expansion mismatch displacements and the separation

of the plastic from the elastic strains.

For loading conditions of this character, it is possible that

high-cycle fatigue behavior may be observed.

A-3.6 CAVEAT 4 — Local Expansion Mismatch

For applications for which the global thermal expansion

mismatch is very small, e.g. ceramic-on-ceramic or silicon-

on-silicon (ﬂip-chip solder joints), the local thermal expan-

sion mismatch becomes the primary cause of fatigue dam-

age. Equation A-4 does not address the local thermal

expansion mismatch. This reliability problem needs to be

assessed using an interfacial stress analysis [Ref. A-9: 35]

and appropriate accelerated testing.

For leaded surface mount components with lead materials

that have CTEs signiﬁcantly lower than copper alloy mate-

rials, e.g., Kovar

or Alloy 42, the results from Eqs. A-1

and A-2 will be optimistic, since the fatigue damage con-

tributions from the solder/lead material CTE-mismatch, the

local thermal expansion mismatch, are not included.

It has shown that the interfacial stresses resulting from the

local expansion mismatch follow [Ref. A-9: 35].

τ∝L (α

Solder

−α

Basr

)(T

max

− T

min

)

[Eq. A-6]

Figure A−2 Solder Joint Pull Strengths for Gullwing Leads Consisting of Alloy 42 from Different Vendors and Copper

[Ref. A-9: 31]

July 1996 IPC-D-279