Assembled board testing Failures

While we tried to analyze test failure cycles that cover a wide range of conditions, components, and board hnishes, the analysis herein is far from exhaustive. In future work, we intend to add new data as to further exploit the correlations presented in this study and to possibly identify new parametric trends. Using two simple metrics to assess progress in SAC or lead-free reliability studies, our estimate is that the industry know-how on SAC assembly reliability is 10 to 24% up the learning curve compared to the Sn-Pb reliability knowledge base. This rough estimate is based on soft data given in Table 1 ... 

In one of the leading UK pharmaceutical companies, a printed circuit board assembly in a plant controller was replaced after a failure with an ostensibly identical unit. To everyone s complete surprise, this simple action, regarded as relatively routine, led to a major operational incident. The cause was found to be an EPROM chip on the replacement board. This chip not only carried an upgraded version of the software on the original but also possessed different operational parameters. Despite the fact that the replacement printed circuit assembly had been successfully tested (in accordance with quality procedures), it was the lack of effective configuration management that allowed the change in the software to elude detection, which led directly to the failure. 

Bias is frequently added for testing of electronic devices, printed wiring boards, and assemblies of electronic equipment. The 85°C, 85 % RH, bias test has been the predominant one in electronics for many years [8], While it sometimes misses failure mechanisms that later occur in the field, it also finds many weak points in new products. It is especially useful for quality control of seasoned devices for which long-term reliability is known to be high if the product passes this test. There are many commercial suppliers of temperature/humidity/bias test chambers and software is widely available to automate the operation, data collection, and data interpretation. Attention to data management is mandatory when hundreds of devices are tested simultaneously. This is frequently required in electronics to obtain sufficient data to make statistically valid predictions of lifetime and failure rate under use conditions. 

Whether these corrosive gas tests are reaUstic for materials other than those used for connectors or for operating electronic equipment is not clear. The test should be carried out, but the observation of no failures should not be taken to mean there will be no field failures in typical urban environments. Similarly, any failures that are observed should be carefully evaluated to ensure that the same mechanism would he operative in field situations. Connectors tire a somewhat unique part of an electronic assembly in that the active part is frequently a noble mettil and the sensitivity of the mated surfaces to failure may be lower thtin many other parts of electronic assemblies. Most failures in electronic assemblies attributable to the environment are due to ionic particle contamination in conjunction with atmospheric moisture. In 20 years of evaluating field failures in the United States, the author has never seen a failure that could be attributed to the effects of SOj, has seen a few caused by H2S or HCl, has heard of a few caused by NOx, and has seen several hundred that were caused by ionic contamination. Clearly, valid accelerated testing of electronic components, circuit boards, and assemblies must include ionic contamination. Emerging methods are discussed in the Fine Particle Testing section in this chapter. 

Mechanically Induced Failures. PCBs may be mechanically loaded by test fixtures or processing equipment, when PCAs are loaded into card cages or fixtured into place with brackets, or when the assembly experiences mechanical shock or vibration in use. In general, once the PCA has been assembled, the interconnects to the components are the weak links in mechanical loading situations, not the boards themselves. 

Reliability of electronic assemblies is a complex subject.This chapter has touched on only one aspect of the problem understanding the primary failure mechanisms of printed circuit boards and the interconnects between these boards and the electronic components mounted on them. This approach provides the basis for analyzing the impact of design and materials choices and manufacturing processes on printed circuit assembly reliabihty. It also provides the foundation for developing accelerated testing schemes to determine reliability. It is hoped that the fundamental approach will enable the reader to apply this methodology to new problems not yet addressed in mainstream literature. 

FIGURE 59.13 Data set comparing the torce-to-tailure rate of Pb/Sn, Pb/Sn/Cu and Sn/Ag/Cu assemblies tor two different smlace finishes (ENIG and SOC). The histograms show a comparison of the different failure modes obtained in the bend tests. All samples were tested within one day of board assembly. (Reprinted with permission from Kyocera SLC Technologies. / [2005J IEEE.)... 

In the next phase of the transition to lead-free technology, lead-free components assembled with Sn-Pb paste will eventually be assembled with SAC paste. Relevant test data (Ref 9, 28, 29, 31) for area array components with SAC balls is shown in Fig. 10, where cycles-to-failure for SAC paste assemblies is plotted versus cy-cles-to-failure for Sn-Pb paste assemblies. The presentation of the data is similar to that of Fig. 9 with symbols and labels identifying temperature profiles, components, and board finishes. All seven data points in Fig. 10 are close to or above the main diagonal, suggesting that during the transition from mixed assemblies (SAC balls/Sn-Pb paste) to 100% lead-free boards (SAC balls/SAC paste), reliability concerns are likely minimized. 

Area Array Assemblies with Sn-Pb Balls and SAC or Sn-Pb Paste. Another scenario of interest during the transition to lead-free technology is that of conventional area-array components using Sn-Pb balls assembled with SAC paste. This scenario is often described as a forward compatibility situation. In Fig. 11, we show cycles-to-1% failure for Sn-Pb ball area array components assembled with SAC paste versus cycles-to-1% failure for similar components assembled with Sn-Pb paste. The data was gathered from relevant test cells in several independent studies (Ref 3,11,12, 22, 29, 32, 33). Figure 11(a) shows the data for assemblies that were cycled between — 40 and 125 °C (— 40 and 257 °F) (6 data points). Figure 11(b) shows similar test data for thermal cycling under milder conditions 0 to 100 °C (32 to 212 °F) (7 data points) and 15 to 95 °C (59 to 203 °F) (2 data points for 144 Input/Output PEG As assemblies with Ni-Au or Sn-Cu HASL board finish). 

For large boards, excessive board flexure, during PWB manufacturing, testing, assembly, and use, often can cause solder interconnect failures. Establishment of the strain rate-strain limit for lead-free solder interconnection on different PWB surface finishes can greatly help safeguard the reliability of the products (Ref 23). 

On the RTV-SM assembly, component location and position had a strong effect on time-to-first-failure for PLCC-84 and LCCC-44 devices. While there was a wide range of performance exhibited across the eight solder alloys, a similarly wide variation in performance was observed among the various component sites on the same board. Furthermore, a wide variation in performance was also observed at all locations across the three replicate boards used to test each solder. It was concluded that this position effect, coupled with general data variability, overshadowed the effect attributable to the solder alloy alone, and masked the ability to distinguish one solder alloy from any other. None of the Pb-free alloys exhibited catastrophic failure or distinguishably poorer performance than the other Pb-free alloys or the eutectic Sn-Pb control. Component type determined the failure time and mode. There were no clear differences between solder alloys or PWB surface finishes under the vibration conditions studied. 

Table 45 lists the Weibull parameters for the 80 lead UTQFP component. Notice that only a few components were tested in some cases. This was a result of large fallout during the board assembly process resulting in misplaced components as well as solder joint bridging. A relative comparison based on alloy A1 is shown in Figure 33. Weibull analysis was not performed for alloy All as there were only two failures for this alloy. Comparing the mean life, all alloys performed better than alloy A1 for this component. Although the mean life for All was not calculated, it would have exhibited a longer mean life compared to alloy A1 based on the first failure data given in Table 45. The same trend can be observed by using a first failure criterion, except for A14, which had lower first failure life than Al.

---