Solder Crack Measurement

Repeated changes in service temperature, due to the weather or engine operation, can cause thermally induced stresses to fatigue crack the solder joints of automotive PCBAs resulting in circuit failure.  Circuit boards are designed to minimize solder fatigue, but in order to verify that PCBAs can survive many years of service use, circuit boards are tested by thermal cycling with the intention of exceeding the number of thermal cycles anticipated for the life of the vehicle.  At the end of the thermal cycle testing the circuit must still be operational and, for added security, any solder cracks that have developed must be less than a stated percentage, often 25%, of the solder joint width.

Because this 25% crack length limit creates a hard divide between pass and fail, or between acceptance and rejection, the method of crack length measurement must be sensible and agreeable to all organizations involved.  While everybody accepts the definition of crack percentage as crack length divided by solder joint width, there is  no universally accepted definition of what crack length or joint width actually mean or how they are measured.  What seems a simple concept in the abstract becomes confused by the details of implementation.

We will discuss some of the issues that create confusion or conflict; we will then discuss how Elmet measures crack length in the absence of contrary customer instructions. We present images of cracked solder joints and present possible interpretations of measurement. 

We will start with an easy one. Consider the solder ball shown in Figure 1.  A crack extends from one side of the solder ball to the other, this solder ball is clearly 100% cracked.  Accepting this obvious statement has implications for how we define the width of the solder joint and how we define the length of the crack.

Figure 1

Since we agree this is 100% cracked, implicitly we accept that crack length is not always measured in a straight line; this crack turns the corner at the edge of the solder pad. Crack length includes the length along the top of the solder pad plus the lengths down the edges of the solder pad. 

Likewise, we must accept that the “width” of the solder joint, in this case the solder ball, is not the width at the centerline or even at the top of the solder pad.  The width is not even a straight line; if it were, the contorted crack length could exceed the length of the solder joint, or crack length could be measured as less than joint width even though the crack clearly splits the joint in two.

The “width” must be defined along the same path as the crack, extending across the top and down the sides of the solder pad. That is the only definition of joint width that leads to (crack length divided by joint width) = 100%.

What if the crack is less than 100%?  The crack in the solder ball shown below extends between A and C - but it does not extend all the way across the solder ball (to simplify the discussion we will ignore the small dark pores at the right end of the solder pad).

Figure 2

What is the crack length?  Consistent with the previous example we would measure the length along the left edge of the solder pad from A to B, written as A-B, and add the length along the top edge of the solder pad from B to C, written as B-C.

Now, what is the solder joint width?  Consistent with the previous example the solder joint width measurement follows the same path as the crack from A to B to C.  We reasonably extend the measurement to D, but what next?  Do we continue measuring horizontally to the right edge of the solder ball or do we measure down the right edge of the solder pad?

Measuring down along the right edge of the pad is consistent with the previous example, AND this distance is slightly shorter than extending the measurement horizontally to the surface at right, so reasoning argues for measuring downward. The crack percentage is given by (A-B-C)/(A-B-C-D-E).

Now consider the QFN solder joint shown in Figure 3. What is the solder joint width? It makes no sense to measure the width of solder along the solder pad because much shorter paths are available for crack propagation. In fact we can see a crack along the vertical face of the QFN lug. To reach 100% cracking, the existing crack must propagate horizontally beneath the lug, so we define solder joint width as the existing crack length plus the horizontal length beneath the lug. Crack percentage is (A-B)/(A-B-C). Using these dimensions the solder joint is 28.1% cracked.

Figure 3

In all three of the examples so far the solder joint width was determined by considering the crack path that is observed.  That is, the crack path always lies along the line that defines joint width.  This is the only way to assure that edge to edge cracking results in a calculated 100% crack percentage.

These examples emphasize that the joint “width” that we are interested in is the distance that a crack would follow, so we must consider the probable crack path before we decide how to measure the solder joint width.

Look at the solder joint in Figure 4. The crack extends from A to A’ at toe of the gull-wing lead.  If this crack propagates further, it will reach the top surface at B’.  The distance from A’ to B’ is much shorter than any other path option, and if the solder joint is to fail, this will certainly be the path of the crack.  Therefore we know the length from A’ to B’, the shortest distance from the crack tip to the surface, is part of the solder joint dimension.

Figure 4

There are three options for measuring joint width to the left of the crack.  We can measure along the gull-wing lead from A to B to C where we reach the surface, or measure from the left end of the crack to the nearest surface at E, or lastly we can measure from A to B to D, connecting the existing cracks or voids at B and D.   

There are two reasons to choose the last option. First, those two small cracks at B and D suggest that the stress field favors cracking along that path, and second, connecting these cracks represents less crack growth than propagating from point A to point E. (distance A-E is 1.16 times the distance A-B plus B-D)  Elmet would measure the solder joint length from B' to A', and also from A to B to D.  The distance from A’ to A is the crack length and is also included in the solder joint width.

In general, the solder joint width is measured as the length of the anticipated crack path. In all three cases so far, the anticipated crack path is the shortest distance from the crack tip to a free surface.  This is usually the case, except when other considerations intervene, such as voids or other crack segments.

Figure 5 shows a solder joint that is 100% cracked with complete separation along the path A-B-C. Considering the crack traverses a large void, what is the crack length and the solder joint width in this example? If the void dimension is included in the crack length but not in the solder joint width, the calculated crack percentage will exceed 100%. If the void dimension is included in the solder joint width but not the crack length, the crack percentage will be less than 100% even though the solder joint clearly has failed and separated. Therefore, to be sensible and to provide meaningful information about the condition of the solder joint, crack percentages must be calculated either with the void included in both the crack length AND in the joint width, or excluded from both.

Figure 5

Crack percentage numbers are intended to reveal something about solder cracking, not the amount of voiding, so typically void dimensions are excluded from the crack length measurement and from solder joint width. The crack percentage of the solder joint in Figure 5 is calculated as (A-B)/(A-B)=100%. The void dimension B-C is not included in either the numerator or denominator.

Figure 6 shows another example. A small crack is visible at the center of the solder joint and the yellow line, labeled A-B, indicates the anticipated crack path.  Length A-B includes a significant amount of voids and a small length of crack.  If we included the voids in the calculations this solder joint would be reported as about 15% cracked.  Excluding void dimensions from the calculations gives the more reasonable interpretation that this solder joint is about 25% cracked.  Voids can have a large influence on crack percentage calculations and, in order to make crack percentages sensible, the void dimensions must be excluded from crack percentage calculations.

Figure 6

Crack percentage measurement is fraught with debate. Disagreement usually arises because a participant has not considered the implications of their assumptions. We hope the above discussion provides some clarity, but comments and other opinions are welcomed in the comment section below.

Warranty Returns, NTFs, and Recall Campaigns

Our proprietary mounting and polishing process gives us the ability to target the same component or set of components on the same board in large numbers, repeatedly, quickly and with great accuracy. In the past, this has been a valuable skill to several of our customers during warranty or recall campaigns, and it could be to you too!

We have taken part in the warranty process at several stages. In some cases, parts have come back from the field with a known failure mode and the OEM and supplier have agreed that the results of cross-sectioning a particular solder joint, screw, weld, axel, through hole or trace will be used to determine a liability split, and we have been contracted as a third party to perform the work. We have also inspected features via X-Ray, used Scanning Electron Microscopy (SEM) to diagnose failure modes and causes, or provided high magnification photographs to identify changes or variations in component manufacture. We can present you and your customer or buyer with the raw data, or you can supply us with pass/fail rules to make the determination.

How does this work in the real world? Suppose that you are supplying a controller for a car’s air conditioning system, and you are getting many warranty returns where the blower does not turn on. The vehicle OEM has collaborated with you and the blower motor supplier and found that this issue can be caused by a particular cracked solder joint on your controller board, or a stuck motor brush. You and the vehicle OEM have agreed to a scheme where Elmet will cross section the suspect joint on a sample of 100 boards returned for this issue, measure the length of any cracks in terms of a percentage, and return to you and the OEM all results and supporting images. From there you may agree to pay warranty or service costs related to the population with greater than 50% crack length, or only those that are fully cracked, for example.

Maybe you discovered that your plant has been using PCBAs from a supplier who operated for a month with an incorrect thermal reflow profile and you want to determine if this has changed the properties of the solder joints. We could cross section dozens, even hundreds, of suspect boards for comparison with your validation artifacts to help you determine your exposure to risk.

On occasion a customer will come to us with a single circuit board for which they have narrowed the problem source down to a single component or area of the board. Or sometimes a customer has a small number of “No Trouble Found” or “intermittent operation” parts that have been returned. We can inspect questionable boards, areas of a board or specific components visually or via X-Ray or SEM for solder issues, tin whiskers or other manufacturing defects to supplement your normal warranty work.

Gear Failures

Gears are one of the most common and fundamental forms of mechanical power transmission, and gear design and manufacture are highly refined. But because gears are so common, and because they are always contained deep within the host equipment, gear failures are both frequent and expensive. 

Gear failure analysis is more detail oriented than many other types of failure analyses because the failure usually cascades through multiple teeth and sometimes through multiple gears.  As an example, teeth that break off by fatigue cracking can be entrained in the mesh and cause overload failure of other teeth.  Anybody that has opened a failed gear box and removed handfuls of fractured teeth and metallic debris knows how extensive the damage can be.

During the failure analysis, the analyst must collect enough evidence to reconstruct the sequence of failure at least to the extent of identifying causal and consequential fractures.  The fracture mode and crack initiation sites of all broken teeth must be documented.  The intact teeth must be examined for evidence of surface wear, sliding contact, tooth tip interference, and other features that might indicate the condition of use of the gear.  This is critical because correctly identifying fatigue cracking but failing to recognize the fatigue was caused by tooth tip interference will obscure the corrective action.

All of the preliminary inspection must be completed before the gear is cut up for the metallurgical inspection.   The roots are inspected for cracks, the pitch line is inspected for frosting or spalling and evidence of sliding contact, the wear pattern is observed, etc.  We attempt to make sense of the observations and interpret them for the customer.  For example, gear teeth with little or no wear show that the gear was not heavily loaded.  If it failed by overload, an unusual event is likely the cause, not the normal conditions of use. 

Failure Analysis of Springs

Springs are one of the most revealing types of failure analysis because the state of stress in a spring is well defined.  The geometry and use of springs strictly limits the mode and direction of applied loading, and the mathematics that describe resulting stresses are well developed.  The location and magnitude of maximum stress is therefore known and typically correlates with the crack initiation site.  We examine the fracture surfaces (fractography) of springs to locate the crack initiation site.  If this site does not correspond to the point of maximum stress, it is likely that material defects or other factors contributed to crack initiation.  We inspect the crack initiation site at high magnification (using scanning electron microscopes) to look for inclusions or laps in the steel, or for damage to the spring surface, that facilitated crack initiation.

If crack initiation did occur at the site of maximum stress, material factors may or may not have been involved.  We must still verify the material microstructure, hardness and, possibly, the chemistry before ruling out material as a factor contributing to failure.  We metallographically prepare a cross section of the spring to inspect the microstructure and measure the hardness as a function of depth.  Hardness outside of the specified range of course contributes to failure, but bulk hardness can be within spec while surface hardness is low, and this still leads to fatigue failure.  Low surface hardness is often caused by decarburization of the spring wire, which we detect metallographically.

Other factors can be derived from the failure analysis.  If crack initiation occurred at the location of peak stress and there was no material factor contributing to crack initiation, then the spring performed as intended and the conditions of use must be questioned.  Was the loading greater than expected?  Where there more stress cycles than expected?

If crack initiation occurred other than at the peak stress location, and there was no material contribution to crack initiation, then the possibility of improper use must be considered.  Essentially this is saying that the stresses must not have been the normal stresses observed in a spring. 

Induction Heat Treating

Elmet’s principle, Arthur Griebel, has worked in the induction industry since 1979.   Working as the Chief Metallurgist at Tocco (now Ajax Tocco) and then as an independent consultant, Arthur has instructed numerous classes in induction heat treating for SAE, ASM, and the Center for Induction Technology.  Arthur contributed to the ASM Handbook on Induction Heating and Induction Heat Treating, and holds three patents relating to induction technology.

 
There are several points in the process of adopting and using induction heating at which outside assistance can be beneficial.  The first is the decision to use, or not to use, induction heating.  Equipment manufacturers can be very helpful, but it can be difficult to get their attention, and it is difficult to know how well they are protecting your interests. The purpose of outside help is to give you basic information and the advantages and disadvantages of induction, an understanding of the process, and the equipment requirements.    

 
If inductors were supplied by the OEM, he will likely give assistance with developing the power levels, heat time, or speeds to obtain the desired heating pattern. Customers that are left to do this on their own sometimes run into trouble, and expert assistance can help tremendously here.  Coil builders are a great source of knowledge, especially if you are just getting started, as they can direct you toward the correct inductor and process design. If you find yourself spinning your wheels, incrementally making changes that get you out of one problem but into another, and you feel a more comprehensive view of the operation is needed, you might benefit from bringing in a consultant. 

Microstructure and the Origins of Metallography

Metallography is the preparation of materials in order to observe the internal microscopic features of the material structure.  The name derives from the inspection of the internal microscopic features of metals, but similar processes are used to inspect the internal structure of any material, including circuit boards, ceramics, composites, and minerals.  In fact, the process was initiated by geologists who polished rock specimens to better observe the individual crystals of minerals.  In 1863 a geologist named Henry Clifton Sorby was the first to polish metal for microscopic inspection.  His father had left him a cutting tool manufacturing shop in Sheffield, England and he was suffering financially because the factors controlling the hardness and toughness of the tools were not understood.  Henry observed that cutting tools sometimes broke with a crystalline appearance similar to rocks.  He suspected that, just like rocks, the microscopic features of metals might be revealed by polishing.  He also suspected that, just like rocks, the properties of the metal might be correlated to the features revealed under the microscope.  This turned out to be the most important discovery of modern metallurgy.

The internal microscopic features of materials are now called the “microstructure” of the material, and the microstructure correlates with properties better than any other material characteristic.  This surprises some people who assume chemistry or heat treatment or manufacturing process has the most significant influence on properties, but steel (for example) can be made very hard despite enormous variation in composition and heat treatment.  The only requirement is that the combination of composition and heat treatment produce a particular microstructure (martensite).

Examining the microstructure of a material is therefore a great way to interpret the properties or the behavior of the material.  Furthermore, as a material undergoes changes due to deformation, heating, or fatigue in service, examining the microstructure is a great way to interpret what has happened or to diagnose a problem.  The microstructure records the history of the material more faithfully than any other characteristic.

In order to see the microstructure, we take a sample through an aptly named process called "metallographic preparation." The intent of metallographic preparation is to accurately reveal the microstructure of the material.  This is achieved by grinding the surface of a sample through successively finer grits. Each grit must remove the coarse scratches as well as the subsurface damage left by the proceeding grit.  Through successive grits the damage to the microstructure becomes finer and finer and the revealed microstructure more and more closely resembles the true microstructure of the material.  The final polishing stages are performed with sub-micrometer size abrasives held in a slurry on a soft cloth so that there is absolutely minimal damage to the microstructure.  The pH of the slurry may be adjusted to chemically remove the final layer of disturbed metal.

While conceptually simple, the process is difficult in practice.  The microstructures of most materials are not uniformly hard and therefore the microstructural constituents polish at different rates.  This produces an uneven polished surface that does not present a single focal plane, and the magnified image of the microstructure will be blurred.  Circuit boards are an extreme example of this because they frequently contain ceramics, such as alumina (aluminum oxide), with a hardness of about 2700 HV adjacent to tin solder with a hardness of about 27 HV.