Assessing Component Damage, Failure Risk

To determine whether lots of plastic-encapsulated microcircuits (PEMs) will fail during service, they are first acoustically imaged in the NXP laboratory to find internal packaging defects, such as cracks. The components are then baked, soaked, put through reflow three times, electrically tested, acoustically imaged a second time, and physically sectioned. Additional tests may be applied when they are appropriate. We have tested thousands of component types, including variations on plastic quad flat pack (PQFPs), quad flat no lead packages (QFNs), ball grid arrays (BGAs) and other well-known PEM designs.

The purpose of testing and imaging after reflow is to determine the moisture sensitivity level (MSL), according to IPC/JEDEC standards, that will be assigned to each part. The data that is gathered also gives a good overall view of the relationship between the internal damage in a component and the component's probable reliability.

For reflow, the PEMs are placed on, but not bonded to, a bare circuit board with no pads or traces. The board is run through reflow three times to simulate a board that is reflowed twice (because it contains both lead-free and lead-containing components, for example), and that later undergoes rework. The parts are not bonded to the board because they must be removed for later tests.

Internal damage visible to acoustic microscopes frequently comes from moisture that has been absorbed by the PEM from the atmosphere. During reflow, the moisture flashes into steam and, in doing so, expands its volume by a factor of ~1600. The resulting pressure may be relieved by a crack inside the part, and the crack may cause the part to electrically fail. Electrical failure may be immediate or eventual, and may be brought about directly by a crack, or less directly by crack-related corrosion. The primary goal of testing is to establish a passing level for each part type — for example, a part that has cracks at Level 2 (floor life of one year) but not at Level 2a (floor life of four weeks).

But testing also frequently indicates how much damage, and what kind of damage, a given PEM can endure before it electrically fails. After the three reflow cycles, electrical testing of each part is performed first because electrical testing will identify those components that have already suffered sufficient internal mechanical damage to cause an open or short. Electrical testing is not directly useful in assigning a MSL to a part because electrical testing by itself does not describe the probable cause of the failure. Next, all parts, including those that fail electrical testing, are acoustically imaged. For the failed parts, acoustic imaging helps pinpoint the cause of the electrical failure. For PEMs that passed electrical testing, acoustic imaging reveals non-lethal damage that occurred during reflow. This damage can be quite extensive without resulting in immediate electrical failure. The components that fail electrical tests generally exhibit massive internal damage that would cause them to be rejected according to the failure criteria in J-STD-020D, but a good many PEMs that pass electrical testing have similar internal damage.

Acoustic micro imaging reveals internal features, such as leadframe dimensions, because its ultrasound is reflected only by material interfaces — molding compound to the leadframe, for example, or die to die attach material. There are no reflections (or return echoes) from homogeneous materials. The most highly reflective internal features are gaps between materials. Delaminations, cracks and voids — the internal anomalies that lead to electrical defects — reflect virtually all of the ultrasound, even if they are as thin as 0.01 μm.

The most common internal damage acoustically seen is some form of popcorn crack. Figure 1 shows the acoustic top view and cross-section of a component in which a popcorn crack (two diagonal lines) originating near the die attach material (red line at bottom) has moved toward the back side of the package. Moisture absorbed through the molding compound tends to wind up on an internal surface, and the biggest surface in a leadframe package is typically the die paddle. In some instances, the die attach material itself may also collect moisture. Changing the die attach material to reduce water absorption is sometimes successful.

1. Top (above) and cross-sectional (below) acoustic image of a popcorn crack (two diagonal lines) extending toward the backside of a plastic IC package. In the sectional view, only the portion of the crack at left reached the package surface.

During reflow, excess moisture on the die paddle expands and creates a popcorn crack. Popcorn cracking generally travels either downward, toward the bottom of the package, or upward. The vast majority of components tested for MSL have popcorn cracks that travel downward. In one respect, a downward crack is less damaging in that there are no wires that the crack can sever. On the other hand, a crack leading to the bottom of the package is something that an assembler will probably never notice, so the crack becomes a quick route for moisture and contaminants to enter the package and cause long-term damage via corrosion. The relatively few popcorn cracks that travel upward always, or nearly always, cause immediate electrical failure by breaking wires or wire bonds. Sonoscan has noted that popcorn cracks in plastic ball grid array (PBGA) packages (sometimes called "popcorn delaminations") travel to the side along the interface between the molding compound and substrate. Although the term "popcorn delamination" may sound odd, it is consistent with two definitions recently defined by IPC/JEDEC: A delamination is an interfacial separation between two materials that are intended to be bonded, while a crack is a separation within a bulk material.

A crack may begin between the die attach and either the chip or leadframe. Such a crack may be limited to the die attach or expand to become a popcorn crack. A crack involving the die attach generally originates between the die attach and die paddle because the die attach material usually adheres more strongly to the die.

Because the die paddle presents a large material interface that can collect moisture, designs that put a die on a paddle that is not a great deal larger than the die seem to avoid cracking because moisture is less likely to accumulate and cause a crack.

2. The dimples in this die paddle, viewed acoustically from the backside, achieved a good bond with the molding compound, but a crack formed beside the paddle.

Even among variations of a single package type — PQFPs, for example — where the basic construction is the same, there are differences in damage response. Often the difference is caused by differences in material properties. One molding compound might be much more crack-resistant than another molding compound. Two packages with very similar designs might therefore wind up with different MSLs. On the other hand, if there are lead-free and lead-containing versions of the same part, the two versions today probably are assigned to the same MSL. In the early days of lead-free parts, they were more likely to be in different levels.

Materials characteristics can also cause problems with adhesion. A down-bond running from the die down to the pad is generally attached by a stitch bond to a silver-plated area on the leadframe. Molding compounds do not always adhere well to silver, with the result that a small local crack is likely to form and the bond can be broken. When necessary for MSL testing, the package is decapped to check the stitch bond for a crack or an open.

In the business of establishing the MSLs for thousands of types of plastic-packaged ICs, the standard electrical test that is used tells whether the device functions or not, but does not give details of why it may have failed. What is more interesting are the packages that pass the electrical test — meaning simply that there is no electrical open or short at the time of testing — but that then show internal damage when acoustically imaged.

When acoustic imaging shows that a number of parts in a given lot have internal defects, the defects may be similar to each other — a delamination along a particular lead finger, for example, or a delamination across the die face. The precise form that a defect takes likely depends on the package size, pad shape and size, and other parameters specific to the part type. The acoustic images give a good idea where to section the parts to make the defects most visible.

The defect criteria that accompany MSLs in IPC/JEDEC J-STD020D are based on acoustic imaging of defects and the results of reliability tests. Section 6.1.e notes that a device is considered a failure if it has an "internal crack extending more than two-thirds the distance from any internal feature to the outside of the package."

3. Moving laterally rather than upward or downward, this popcorn crack (red and yellow) followed the plane of the lead fingers.

But acoustic imaging has also revealed large cracks that are clearly not delaminations that follow the horizontal plane of the lead fingers. Such cracks typically encompass several lead fingers and the molding compound between them. The cracks can be acoustically distinguished because they lie just above or below the lead fingers and, if on top, may mask the individual lead fingers.

In the top-view acoustic image of the PQFP (Fig. 3), the crack has followed the lead fingers. In places, the crack has opened above the tape (red areas on tape) and, in other places, it has passed under the tape (white areas). It has followed some lead fingers to the package exterior.