New study reveals component defect levels: an extensive study of almost one billion solder joints reveals the defect levels and most common defects for components on PCBAs.

In 1999, a study was conducted of certain production data from 15 different companies in Europe and North America; eight original equipment manufacturers (OEMs) and seven electronics manufacturing services (EMS) providers were included. All of the companies used automatic x-ray inspection (AXI) to examine a combined total of more than one billion solder joints. (1)

Using the same data source, a recent study examined the defect levels for different component types such as ball grid arrays (BGAs), quad flat packs (QFPs), J-lead components, connectors, chip resistors and chip capacitors. This new study also analyzed defect levels for different pitch components and different size components.

The data were collected automatically at the companies' facilities using the same AXI system and special software programs. The data collection method worked as follows. When a printed circuit board assembly (PCBA) was inspected by the AXI system, a file was created with inspection results. The software parsed this file and extracted information about the PCBA, such as when it was tested, serial number, board type, number of solder joints inspected and number of joints detected as defective. This information was then automatically written into a Microsoft Access database.

The second important event occurred when a defective PCBA was repaired. At the repair step, more information was inserted by the repair operator and an additional file was generated. The new file contained information with serial number, board type, when repaired, repair operator name and, for each defect, component name, pin identification, 'defect type and if the defect was repaired. When this file was generated, the software again parsed the information from the file and wrote it into the Access database.

The total number of boards included in the 1999 study was 328,964, the total number of solder joints was 1,013,770,241 and the total number of different board types was 566. For the new study, one additional computer-aided design (CAD) file for every board type was needed. However, this file was only available for 393 board types. So the data in the new study is slightly smaller than the original; it includes 253,018 boards inspected, 843,171,778 solder joints inspected, and 910,111 defects that were confirmed by the repair operator.

Defect Levels

A defect is an unacceptable deviation from a norm. In this study, defect levels are reported as parts per million (ppm), and the ppm levels are reported on a solder joint basis. Every solder joint is one defect opportunity, so the choice is either zero defect or one defect per solder joint.

For all inspections done by the AXI system, the repair operator can classify the defect into three categories:

* R--the solder joint is defective and repaired, or a component placement defect was repaired.

* N--a borderline defect, not significant enough for a repair, but clearly a process deviation worth noting and potentially corrected by process engineers if enough of these defects occur.

* F--questionable defects, outside the scope of this study.

The N classification was not used consistently by all companies; some used it to mean this defect should be repaired later at a special repair station. Because of this inconsistency, the R group and the N group were added together and shown as defects. The result may be defect levels that appear slightly higher than they actually are. However, defects detected at other test steps, such as in-circuit test, functional test and system test, are not included, which may lower the apparent defect level. For the 15 different companies, the average defect level was 1,083 ppm (R + N).

Additional CAD Data

The AXI system inspects different solder joint types using different algorithms. The algorithm used for a specific component's solder joint is specified in the CAD file. The CAD file also contains x, y, width and length information for each solder joint and each component. By parsing the CAD file, the database was updated with that information.

In addition, almost all companies used the letter U for the first letter in the designator name for an IC, the letter J for a connector, R for a resistor, etc. This information was also added into the database. With this new updated database, new queries could be written and additional information retrieved.

Defects per Joint Type

This information revealed that the most common joint types in the data sample were gullwing, with around 396 million joints, followed by joints at chip components, with around 273 million joints. In third place were BGA joints, with around 90 million joints. The two other major joint type groups, in decreasing order, were J-lead and plated through holes (PTH). The joints in these five groups comprise about 99 percent of the joints in this database, so only these groups will be discussed further.

Figure 1 shows the defect levels for these joint types. The PTH group has the highest defect level at close to 4,000 ppm. This result should not be a surprise because most PTH-pins are soldered in a wave process, which introduces significantly more defects than the reflow process.

[FIGURE 1 OMITTED]

Gullwing has the second most defects on a ppm basis. The gullwing group has the most defects because it has a rather high 1,400 ppm level and has the largest population in this study. The groups chip, J-lead and BGA have similar defect levels of around 600 ppm. The average defect level in this data set is 1,079 ppm.

The fault spectrum for all joint types in this study is shown in Figure 2. The most dominant fault class is opens with 48 percent, followed by shorts with 23 percent, insufficient solder with 15 percent, and missing components with 4 percent. About 25 percent of the defects, including insufficient solder, misaligned, voids, and orientation, are difficult to find with electrical tests, such as in-circuit test, functional test or system test, but they could provide a defect in the field at the end customer.

[FIGURE 2 OMITTED]

Gullwing joints

The gullwing is the most common solder joint type in this data sample. This study includes 396,115,325 gullwing solder joints. Most of these joints are on QFPs, and some are on dual-inline integrated circuits (ICs) and surface-mount connectors. The total number of defects in this group is 547,601, which is a defect level of 1,382 ppm, a little higher than the average.

Figure 3 shows the defect levels for 16 mil, 20 mil, 25 mil and 50 mil pitch between each component pin. This data only includes QFP ICs. As expected, the finer the pitch, the higher the defect level. The 16 mil pitch has a defect level of 13,080 ppm, an order of magnitude higher than the average. The 20 mil pitch has a defect level of 1,878 ppm, which is higher than the average defect level in this study, but significantly lower than for the 16 mil pitch. The 25 mil pitch has a defect level of 950 ppm, slightly lower than the average of 1,079 ppm, and the 50 mil pitch has a defect level of 650 ppm.

[FIGURE 3 OMITTED]

Accordingly, to reduce defect levels, 16 mil and 20 mil QFP components should be avoided if possible. In many cases, you may not be able to select component packages with lower defect levels. However, if the 16 mil or 20 mil QFPs are selected, more defects should be expected.

Figure 4 shows the fault spectrum for all gullwing joints. The four most common defect types are opens with 65 percent, shorts with 16 percent, insufficient solder with 11 percent, and misaligned with 5 percent. This fault spectrum, plus the defect levels in Figure 3, strongly indicates that a very significant part of the defects are coming from pin problems such as coplanarity problems and bent pins.

[FIGURE 4 OMITTED]

Chip joints

The second largest group of solder joints is on chip components such as chip resistors and chip capacitors. This study includes 272,533,954 solder joints in this category, with 174,704 confirmed defects, resulting in a defect level of 641 ppm for this group.

Figure 5 shows defect levels for some different sizes of chip components. For the defect levels of the different gullwing QFP pitches, the result was not surprising. Similarly, one would expect that the defect levels for smaller chip components would be higher. However, that is not the case. The 0402 has 311 ppm, the 0603 has 91 ppm, and the 0805 has 298 ppm. These results are significantly lower than the average defect level.

[FIGURE 5 OMITTED]

For the 1206, the defect level increased to 702 ppm, still below the average. The 2010 has a defect level of 2,848 ppm and the 2512 has 4,893 ppm; significantly higher than the average and around an order of magnitude higher than the smaller chip components. Process engineers interviewed were not able to provide an answer for why this result happened. One engineer suggested that the smaller components may more easily self-correct in the reflow oven, but this explanation is only a theory at this time.

Regarding the fault spectrum for all chip component solder joints, the highest defect type is shorts with 47 percent, followed by opens with 26 percent, unlike the gullwings.

BGA joints

BGA solder joints were 11 percent of the whole study population, which was up from 3 percent in a similar study of three companies in 1998. (2) More and more BGAs are being used on PCBAs, and this data confirms that fact. The study contained 90,409,810 BGA solder joints and 56,044 defects, or 620 ppm.

Regarding defect levels for different pitches between the balls, the smaller the pitch, the higher the defect levels. The group of 50 mil pitch is significantly larger in population, around 81 million joints, as compared to 7.6 million for the 100 mil pitch and 1.7 million for the less than 50 mil pitch.

While the defect levels for BGAs are higher than expected, only 1,309 BGA components had one or more defects. The average number of defects per defective BGA was 43 defective balls. Of the group of 1,309 defective BGA components, only 447 had one or two defective balls, the remaining 862 BGA components had more than two defective solder joints or defective balls. So the overall quality of BGAs is good, even if the defect level on a joint basis is rather high. A key point is that repairing one BGA with one solder joint or ball defect is as difficult as repairing a BGA with 500 defects.

The fault spectrum for the four highest groups of BGA defects is shown in Figure 6. The most dominant defect is insufficient solder with 42 percent. In this study, 11,783 balls had enough voids to be marked as defects and are 21 percent of all BGA defects. The third defect group was shorts at 12 percent.

[FIGURE 6 OMITTED]

J-lead joints

J-lead solder joints are rather easy to manufacture with low defect levels. Including IC packages with J-lead, this study has 48,862,591 J-lead solder joints, with 22,805 defects, resulting in a ppm level of 466.

Regarding the defect spectrum for J-lead devices, the most dominant defects are opens with 74 percent, followed by insufficient solder with 11 percent and shorts with 8 percent. No real surprises occurred in the J-lead data or fault spectrum.

PTH joints

As a group, PTH solder joints have a rather high defect level of 3,976 ppm. Most PTH solder joints are manufactured using the wave process or selective wave process. The wave process is widely known for introducing many defects. In a study of different test/inspection systems' effectiveness, more than 50 percent of all defects were introduced in the wave process.

The fault spectrum of PTH solder joints are shorts with 36 percent, insufficient solder with 35 percent, and missing components with 20 percent (Figure 7). The insufficient defect is, most often, insufficient barrel fill; not enough solder has propagated into the hole. The missing defect is either a totally missing pin or a pin that has not been pushed through the hole.

[FIGURE 7 OMITTED]

Surface-mount and PTH connectors

Except for 16 mil QFP ICs and some chip components, connectors are the component type with the most defects. An analysis of soldered through-hole connectors and surface-mount connectors was done on the database. Through-hole connectors were slightly more popular with 10,479,336 joints as compared to 7,308,288 joints on surface-mount connectors. The defect level for surface-mount connectors was slightly higher with 2,727 ppm as compared to 2,289 ppm for through-hole connectors.

As mentioned, the main problem with through-hole connectors is the wave process, which introduces many defects. The problem with surface-mount connectors is the pins in the connector. The pin material selected must have a good spring feature, which causes problems in the reflow process. Often, these connectors float on top of the solder. The solution is not to push down the connector during reflow, because the melted solder is not flowing properly around the surface-mount connector pins. Because of the problems with each of these connector types, no obvious choice exists to select with the objective of reducing defect levels.

Conclusion

These findings are based on three to six months of production data gathered from 15 different manufacturing facilities in North America and Europe from the fall of 1999. The raw data comprise over 1 billion solder joints on 328,964 x-ray inspected PCBAs. The main conclusions of this study are:

* The average defect level in this study is around 1,I00 ppm, which is significantly higher than what is usually claimed by the industry.

* The most common defects are opens (48 percent), shorts (23 percent), insufficient solder (15 percent) and missing components (4 percent).

* The most common solder joint type is gullwing, followed by solder joints on chip components.

* The use of BGAs was significantly higher in late 1999 as compared to a similar study done in early 1998.

* The 16 mil QFP had the highest defect levels, followed by 2512 and 2010 chip components, and then by surface-mount connectors and through-hole connectors.

This article originally presented at 2001 SMTA International.

References

(1.) Oresjo, S. 2000. Year 1999 Defect Level and Fault Spectrum Study. Proceedings of SMTA International 2000. Edina, MN: Surface Mount Technology Association.

(2.) Oresjo, S. 1999. Test and Inspection as Part of Process Control. Proceedings of SMTA International 1999. Edina, MN: Surface Mount Technology Association.

Stig Oresjo is senior test strategy consultant with Agilent Technologies, Loveland, CO; e-mail: [e-mail].