Acoustic surface image of a BGA, whose surface is domed and marked by local spikes (white) caused by voids between the die and the mold compound

Acoustic surface image of a badly warped plastic BGA. The difference in elevation between the low point (X) and the high point (Y) is 0.4mm

Profile of the BGA surface along the horizontal line indicated by arrows in Figure 1

The same BGA with its surface distortion exaggerated in a 3-D image that has been tilted

Internal acoustic image of the BGA showing voids (arrows) in the bulk of the mold compound

Internal acoustic image of the die attach depth, where red areas are delaminations

Package-on-package (PoP) assemblies, and in particular PoPs in which both members are BGAs, are becoming more widely used. PoPs offer many of the space-saving advantages of stacked die with fewer threats to long-term component reliability. But they are not without their problems, however. Industry studies have shown that assembling the PoP magnifies problems of mismatch in the coefficients of thermal expansion of the various materials involved.

A PoP configuration typically consists of a logic device on the bottom and a memory device (which may be a wire-bonded two-chip stack) on top. Industry studies1 have shown that assembling the PoP magnifies problems of mismatch in the coefficients of thermal expansion of the various materials involved. Mismatch in turn can lead to excessive warping of the BGA package, with the ultimate risk of open solder joints. It has also been shown that the degree of warping in the bottom BGA depends strongly on the ratio between the area of the die and the area of the BGA package. BGAs having relatively large die in comparison to the package size warp more severely, although precise control of the memory BGA on top can limit warpage to acceptable levels. The degree of warpage is also determined by the thickness of the substrate, the material attaching the die to the substrate, and by the thickness and material properties of the mold compound.

A new technique developed by Sonoscan, and for which a patent is pending, uses an acoustic microscope to create an acoustic flatness profile, or topographic map, of the BGA's surface while at the same time imaging internal defects and layer thicknesses within the BGA. The topographic map itself reveals the degree of concave or convex warping and, in conjunction with the acoustic images and data showing internal features and layer thicknesses, gives data that is very useful in evaluating the reliability of the BGA. The topographic map may also show spikes where internal anomalies such as voids have pushed the surface upward locally.

The acoustic flatness image and the internal acoustic image are produced simultaneously because of the way ultrasound behaves in solid materials. As the ultrasonic transducer scans the sample and sends a pulse into the sample at each x-y coordinate, it receives return echo signals from material interfaces, but not from the bulk of homogeneous material. A mold-compound-to-silicon interface might send back an echo of medium amplitude, but the echo from the interface between any solid material and an air-filled gap (crack, delamination, void) sends back an echo of very high amplitude. Data from all of the scanned coordinates is used to make an acoustic image that shows expected internal features such as the die and lead frame, and unexpected high-amplitude features such as defects. The arrival time of the return echoes is also used to measure the depth of a feature, or the thickness of a layer of material. In a PoP, the thickness of the mold compound can be measured. Perhaps more important, the uniformity of the mold compound, as well as some of the material properties of the mold compound, can be measured. Imaging is generally limited, though, to the non-stacked or top BGA because the thin substrate may be a poor transmitter of ultrasound and typically there is no attachment between the bodies of PoP stacked components.

The new surface-mapping technique ignores internal features; instead, it measures, at each of the thousands or millions of coordinates, the distance from the transducer to the top surface of the part. The high-resolution acoustic surface map made in this way shows the topography of the part's surface in color, and is acquired at the same time as the internal acoustic image and desired depth data within the part.

The acoustic surface map was developed because of the advantages it provides when dealing with packages such as plastic BGAs that are susceptible to warping that can place unwanted stresses on connections. Looking at both the surface image and the internal image gives the user much more critical information about the BGA, PQFP or other part. The acoustic image of internal features may show no delaminations, cracks or voids – a flawless part, internally – but the acoustic surface image may show a degree of warping that threatens the interconnect integrity of the PoP structure. Or the part may be essentially flat but have significant internal defects. Comparing the two images makes it much easier to evaluate the true status of the part and its probable long-term reliability.

Figure 1 is the acoustic surface image of a plastic BGA. The range of colors shows that the BGA is somewhat bowl-shaped with the low elevations (red) near the center and the high points (blue) around the perimeter. The image also suggests that the surface distortion is not uniform. The yellow area, for example, extends farther toward the lower left than in other directions, suggesting that the lower left is slightly less curved.

Selecting any two points on the surface image provides the difference in elevation between those two points. In the case of this BGA, the lowest point (marked X) is 0.55 mm from a reference plane. The highest point (marked Y) is 0.15 mm from the reference plane. The total warping, then, is 0.40 mm, or 400 microns – serious enough to be a reliability concern since 400 microns is 4 times the average allowable coplanarity of 100 microns per the JEDEC design guide for PoPs.

The non-uniformity of the surface distortion is shown in Figure 2, a profile of the BGA's surface along a horizontal line drawn through the middle of the BGA package (marked by arrows in Figure 1).

The degree of warping can be made more visible by creating a 3-dimensional model that exaggerates the vertical magnification and by tilting the model, as seen in Figure 3. The grid around the 3-D image and the scale at right indicate the 0.4 mm total warping.

Figures 1, 2 and 3 give detailed information about the BGA's surface. The acoustic images of internal features, made at the same time, reveal problems that are not visible on the surface.

Figure 4 is the acoustic image of the bulk of the mold compound in the same BGA, using echoes down to the level of the die. The irregularly spaced dark features (two of the larger ones are marked by arrows) are voids in the mold compound. Basically, these are large and small air bubbles that were introduced with the fluid mold compound and became trapped when the mold compound cured. But the presence of these voids in the mold compound has not caused any distortion of the surface of the mold compound. If there were distortion, it would be visible as vertical spikes in Figure 1 and especially in Figure 3. The voids are an artifact of injection molding and probably have no connection with the warping of the BGA, but they indicate that the molding compound is not uniformly filled. If the molding compound has any significant density variations, this would also be detected.

Using return echo signals from deeper within the same plastic BGA package – specifically, at the die attach layer – produced the acoustic image shown in Figure 5. The red areas in the die attach are delaminations, which cover a significant percentage of the die attach area and which could act as thermal barriers that would cause the die to overheat and fail. (The two somewhat indistinct black features in this image are the acoustic shadows of two of the small voids in the mold compound some distance above the die. The same voids are also visible in Figure 4). A non-destructive acoustic cross-section of the BGA was also made. It showed that the die is strongly tilted.

To sum up: the combination of acoustic surface imaging and acoustic imaging at specific depths revealed that this plastic BGA: 1) was severely warped, with a maximum altitude variation of 0.4mm (400 microns); 2) had numerous scattered voids in the bulk of the mold compound; 3) had large delaminations in the die attach; and 4) had a badly tilted die.

The starting picture is the acoustic surface image of a plastic BGA with a different set of anomalies. First, the surface of the part is domed rather than concave. Second, the surface has been pushed upward locally in several spots (white features). Acoustic imaging of internal features showed that these surface spikes matched the locations of voids (trapped air bubbles) between the die and the mold compound. The voids were able to distort the surface of the part in an upward direction enough to cause visible spikes in the acoustic surface image.

Making an acoustic surface image has shown to be a useful technique not only for BGAs but for any package where there may be some correlation between internal anomalies and surface distortion, or – in the absence of internal anomalies – to measure the warping of the package to determine whether stresses are so great that they might lead to electrical failure during service. Failure might result from damage to solder balls or the substrate interconnects. Failure might also result from internal cracks caused by warping. Such cracks might break wires, but even if they do not, they are likely to become collection points for moisture and contaminants that will promote corrosion and eventual electrical failure.

Acoustic surface imaging is commonly used on flip chips (many of which tend to be domed) and on parts of any size; even quite small plastic packages can reveal unexpected surface distortions. The scan area can be extended to include the whole area of a board, where the surface image of the board itself may reveal warping that can impact component reliability.

www.sonoscan.com

EPPE 461

Tom Adams, consultant, Sonoscan Inc., Elk Grove Village (Illinois, USA)