Head-on-Pillow Defect Detection: X-Ray Inspection Limitations

This study has its roots in a solder paste qualification test that aimed at finding a low, or zero, halide classified solder paste that could replace a higher halide content one. When using a “weaker” flux in a solder paste, there is a risk that the wetting toward PCB solder lands, leads, and terminations will be negatively affected. Many different wetting tests were, therefore, performed. One of these wetting evaluations was a HoP defect test that followed a method first presented by Murch et al. [1].

The official definition of HoP defect is stated in IPC-7095D [2] as “a solder joint comprised of two metallurgically distinct masses formed from a ball grid array (BGA) ball and reflowed solder paste which has incomplete or no coalescence.” A cross-sectional image of a HoP defect is shown in Fig. 1.

The x-ray inspection of the deliberately induced HoP defects followed by prying off of the components did not only show how difficult it is to detect this failure type, but also showed many different nonideal solder joint shapes that the electronics industry needs to find out how they affect strength and reliability for different packages and working environments.

An x-ray image from this HoP study that shows many different solder joint shapes is presented in Fig. 2.

A HoP defect evaluation method [1] was chosen, in which two chip components are to be placed a few pad rows into the pad site of a test BGA after solder paste had been screen-printed (see Fig. 3).

After the placement of the two chip components, a BGA shall be placed partly on top of the chip components (see Fig. 4).

The idea with this HoP defect evaluation method [1] is that the different rows of BGA solder balls will get in contact with the solder paste flux at different times during the reflow soldering process. Some solder balls will be in contact with the solder paste flux from the very beginning, whereas others will not get in contact until the solder has completely melted and “embraced” the chip components.

The following activities were performed in this evaluation:

1. Measurements of components and boards

2. Preparation of stencil, SMT program, reflow profile, etc.

3. Screen printing of test boards

4. SPI measurements of the screen-printed solder paste deposits

5. Placement of two chip resistors on each test BGA site

6. Placement of BGA on the sites with the two chip resistors

7. Convection reflow soldering in an air atmosphere

8. X-ray inspection and analysis

9. Prying off the BGAs and analysis of results

10. Comparison of the results from the x-ray and prying analyses

Using the afore-described methodology to compare HoP susceptibility of different solder pastes requires extensive knowledge of packages, boards, added materials, and processes. The following must be known and under control:

1. BGA solder ball heights

2. Chip heights

3. BGA solder ball coplanarity

4. Solder paste print heights

5. The distance the components are pressed into the solder paste deposits during placement

By choosing a relatively small BGA component and a symmetrical board layout, the warpage during soldering would in this case have a negligible effect on the result.

In this HoP defect evaluation, a test board originally designed for a BGA ball and BTC/MLF thermal pad solder joint void test was used. For the HoP defect test, 0402 chip resistors and 256-ball BGAs with 1.0 mm pitch were chosen. For the assembly, nine different SAC305, no-clean, type 4 solder pastes (named A to I) were tested.

A two-layer, 2.1-mm-thick test board with ENIG pad finish was used in this test. There were five sites for five different BTC/MLF components and three different BGAs as well as a big PA transistor on each board, but for this HoP defect test, only one BGA site per board was used. Panels with two test boards each were used in this test.

The NSMD PCB pads for the HoP defect BGA assembly test were round with .41 mm diameter and .53-mm solder mask openings.

The test board panel can be seen in Fig. 5.

Standard BGA256 packages and 0402 chip resistors were used in this test.

Images of the test components are shown in Fig. 6.

The BGA package had 1.0 mm pitch, daisy-chain pattern, SAC105 alloy solder balls, and 17 mm side lengths. The BGAs' ball heights varied from .36 mm to .39 mm before soldering.

Measurements of the resistors showed heights between .29 and .32 mm.

The assembly process started with solder paste printing onto the BGA site pads followed by placements of 0402 chip resistors and BGA256 components. Finally, the assembled test boards were reflow soldered in an air atmosphere.

In the screen printing process, a .127-mm (5-mil)-thick, laser cut, stainless steel stencil with .41 mm square apertures with .05 mm corner radii was used.

The paste deposit heights and volumes for each solder deposition and for each solder paste were measured using a production SPI equipment. All tested solder pastes showed stable height and volume distributions for the solder paste deposits on the pads for the BGA sites.

After screen printing and SPI measurements, two 0402 chip resistors were placed four pad-rows into each of the BGA packages' pad sites. The 0402 chip resistors were placed on two BGA256 pad sites for all tested solder pastes, except for paste A, where only one BGA site had chip components. Note that the 0402 chip resistors were rotated 90° compared with the study described in Figs. 2 and 3 [2]. The reason for this was to obtain equal pressure from the BGA balls on both terminations of the chip resistors at the same time, thus minimizing the risk of “tombstoning” the 0402s underneath the BGA.

Assembled chip components on a screen-printed BGA site are shown in Fig. 7.

The 0402 chip resistor placements were manually inspected before the BGA256 packages were placed on top.

An image of a BGA256 placed onto a site with two 0402 chip resistors underneath is shown in Fig. 8.

Separate measurements of placements of the 0402 chip resistors and BGA256 packages showed that the chip components were pressed into the printed solder paste so much that their total height after placement was about .36–.38 mm. This means that from the fifth or sixth row from the left in Fig. 9, the solder balls do not touch the solder paste until the solder paste and BGA solder balls start to melt.

Reflow soldering was performed in a 12-zone convection oven in an air atmosphere using a reflow profile that suited all tested solder pastes. The BGA solder joint temperature profile is as follows:

1. Time above liquidus for the SAC105 solder balls: 75 s

2. Time between 150°C and 220°C: 94 s

3. Max ramp: 2.0 °C/s

4. Peak temperature: 245°C

5. Peak reached after 4 min 0 s

The reflow soldering profile is shown in Fig. 10.

Inspection of the test boards showed good results with correctly placed components, good wetting, and well-formed solder joints. One example of an assembled test board panel, with the HoP defect test packages encircled in dark blue, can be seen in Fig. 11.

The HoP defect package positions are called No. 5 and No. 10 (see Fig. 11).

One post-reflow side view image of a soldered HoP test BGA256 is shown in Fig. 12. Note the difference in standoff between the solder joints to the left, which are furthest away from the preplaced 0402 chip components, and the solder joints to the right, which are closer to the chip components.

All test boards were successfully assembled.

The most common, nondestructive, method to inspect hidden BGA solder joints is to use x-ray inspection.⁷ In this study, a modern 2D/2.5D x-ray equipment with oblique views up to 70° and 360° sample rotation was used (see Fig. 13).

In these BGA solder joint inspections, the x-ray detector was tilted 70° and the inspection sample table was rotated 45°. An example image from one of the x-ray inspections using this view is shown in Fig. 14.

In Fig. 14, the 0402 chip resistors placed underneath the BGA256 are visible as well as the different shapes of the solder joints. The risk of getting solder joints with poor coalescence (and possible HoP defects) between the BGA balls and the applied solder increases toward the lower right of the image. The standoff decreases toward the upper left and the risk for HoP defects decreases in this direction.

The same BGA as shown in Fig. 14 is shown in a top view (0° tilt) image in Fig. 15.

The two 0402 chip resistors are also visible in Fig. 15, and it is clearly shown how the solder joint diameters increase when looking at the solder joints further to the left of the x-ray image, most far away from the chip components. Larger solder joint diameters are a sign of lower standoffs.

When inspecting the x-ray images of the solder joints that had been formed after having placed and soldered BGA256 packages on component sites containing two 0402 chip resistors each, it was necessary to decide exactly which solder joint shapes should be judged as potential HoP defects. Some of the solder joints produced in this test may have been classified as “fractured solder connection” or “waist” defects according to IPC-A-610G Acceptability of Electronic Assemblies [3], but it is, in many cases, difficult to draw the line between these two error types and HoP defects. Therefore, all these three error types were classified as potential HoP defects in this evaluation according to the following criteria.

1. HoP defects:

   • Solder joints with two distinct structures—poor/no coalescence between the added solder paste and BGA ball [1]

   • Solder joints with a “waist” at the lower section of the joint—no full wetting between the added solder paste and the BGA ball [3]

Examples of images with solder joints judged as HoP defects in this x-ray inspection are given in Fig. 16.

Solder joints regarded as good, with no HoP defects, were fully wetted with a single solder joint structure formed between the board pad and component-side BGA pad. These correct BGA solder joints, regarding the HoP failure mode, could be fully collapsed or elongated.

Examples of good solder joints (no HoP) are shown in Fig. 17.

Examples of the best and the worst performing solder pastes in this HoP investigation are shown in Figs. 18 and 19.

As can be seen in Fig. 18, all BGA solder joints have one single structure even though they have different shapes, with elongated solder joints where the standoff is at its highest and more compressed solder joints on the component side furthest away from the 0402 chip resistors. This was the solder paste with the very best result in the HoP evaluation with no HoP defects found.

In Fig. 19, many HoP defects can be found at the BGA solder joint rows with the highest standoff. There is a decreasing number of HoP defects toward the center solder joints of the BGA and no HoP defects at all at the BGA solder joint rows with the smallest standoff from the board. In many of the rows containing HoP defect solder joints, some of the HoP defects are found adjacent to correctly formed solder joints.

In the calculations of HoP failures in this evaluation, 252 (256-4) solder joints per BGA were considered as defect opportunities. The best solder paste had 0% failures (the x-ray image shown in Fig. 18) and the worst 38.9% (the x-ray image shown in Fig. 19). A summary was made, showing the nine different solder pastes' ability to mitigate HoP defects. This summary is given in Fig. 20.

Fig. 20 shows that the HoP mitigation ability differs greatly among the nine evaluated solder pastes. Solder pastes D and H are, e.g., HoP defect free (or close to), whereas solder pastes A and I are the worst performers with around 30%, or even more, of the solder joints expected to have HoP defects.

The x-ray HoP defect judgments were difficult to make and needed to be verified. The components were, therefore, priedoff, and both the pried-off components and the board component sites were then inspected.

To verify the HoP defects in this study, the x-ray–inspected BGAs were pried-off and then both the boards and the packages were inspected.⁸ All pry-offs were made in the same manner, starting from the BGA edge furthest away from the 0402 chip resistor placements.

After having pried-off the components, it was very easy to assess real HoP defects using microscope inspection. When a HoP defect has occurred, the BGA solder ball remains at the package side and has a dimple from the solder on the removed board pad. There is no cracked area on the remaining solder ball, and the ball edges around the dimple are smooth. An example of HoP defects verified by prying off a component is shown in Fig. 21, where eight real HoP defects are verified and four solder joints are found to be strong enough to peel off the motherboard pads.

When prying off correctly formed BGA solder joints, the balls are most often completely removed together with the pads, either from the board or from the package. These solder joints could also break, leaving cracked solder surfaces on both the board and package sides.

An example of two pried-off components, soldered with solder paste I, is shown in Fig. 22.

By prying off the BGA test components, an easy and secure verification of real HoP defects could be performed. This is, however, a destructive method that cannot be used on real products.

When counting the real HoP defects after having pried-off all test packages, a comparison between the previously judged HoP defects found by x-ray could be performed.

A summary of this comparison is given in the graph in Fig. 23.

As can be seen in Fig. 23, the differences between HoP defects found in x-ray inspection and after prying off the components could be large. As the HoP criteria used in this x-ray analysis also included fractured solder connections and “waist” defects, it was expected that slightly more HoP defect registrations should be made when performing the difficult x-ray image judgments compared with the verified HoP defects after prying. A ratio of at least 70–80% correctly verified HoP defects was estimated to be found. However, in many cases, the differences were much bigger than this, which are shown in Table I.

It is obvious that many solder joints that look as real HoP defects in x-ray are not HoP defects in reality. It can also be mentioned that the solder paste with no HoP defect at all was the only ROL1 classified solder paste in this test. All other pastes were classified as ROL0.

Comparisons of x-ray images and microscope images after having pried-off components were performed. One BGA package corner from this comparison is shown in Fig. 24. An explanation to Fig. 24 is that if the x-ray image would be seen from below, the same view as for the pried component image will be obtained. In both images, the red encircled solder joints and the solder joint with a blue square around are the same.

In Fig. 24, all nine BGA solder joints were judged as HoP defects at the x-ray image analysis. However, after prying, only two of them were verified as real HoP defects. The other seven solder joints were strong enough to even pull off the pads and the traces between the pads.

The component corner solder joint within the blue square in both images in Fig. 24 is enlarged in Fig. 25.

In the x-ray image shown in Fig. 24, in some of the solder joints, the applied solder and ball seem to have, at least, partly wetted together, but it is impossible to see any differences between the two real HoP failures and, e.g., the corner solder joint within the blue square (Figs. 24 and 25) when studying the x-ray images of these BGA joints.

The previous sections show that it is almost impossible to guarantee detection of all HoP defects by using only x-ray inspection as detection technology. What is possible to see in x-rays is “HoP-shaped” BGA solder joints, and these are indeed solder joints with a high risk of having much lower strength and reliability than fully wetted, homogenous ones.

In IPC-A-610 revision G, Acceptability of Electronic Assemblies [3], section 8.3.12.3, HoP defects are only shown by a photo, and it is barely possible to visibly see HoP defects other than in peripheral BGA rows, and this standard gives no guidance in interpreting x-ray images. The HoP defect criterion in this standard is “ball is not wetted to solder,” which is easy to agree with. However, a “waist” in a BGA solder connection is also treated as a defect (but not classified as a HoP), and this defect is most often possible to detect with modern x-ray equipment.

The side view images used to show HoP defects and BGA solder joints with “waists” in [3] are given in Fig. 26.

Although the x-ray images cannot clearly show whether the ball and the solder have wetted together, the x-ray images could, most often, show when there is a “waist” in the solder connections, which also could disqualify these solder joints.

This study has shown that the occurrence of HoP, “waist,” and fractured solder connection defects is difficult to distinguish from each other using x-ray inspection. As these errors can occur both along the BGA outer rows and under the central parts of the package, the optical side view inspection criteria stated in IPC-A-610 revision G [3] are not sufficient as quality criteria. Consequently, an x-ray inspection criterion based on treating any “waist” in the BGA solder joints as a defect is recommended for the next revision of IPC-A-610.

The authors are grateful to the following colleagues who have contributed to the development of methods, performed tests and measurements, and reviewed the paper: Anne-Kathrine Knoph, Julia Xu, Rebbeca Dehlin, Dzemila Durakovic, Naser Ismaili, Mats Karlsson, Binas Nisic, and Jesper Wittborn.