News

In this analysis, Pb-free and SnPb TVs showed less than a 10% difference using a common test vehicle design.

Automated x-ray inspection (AXI) can detect most solder-related defects, including shorts, excess solder, misalignment, voids, insufficient solder, open/lifted solder joints and marginally accepted joints, which pass in-circuit and functional test but often result in field returns. Pb-free and SnPb are different materials, and their grey levels, or x-ray images, are different under AXI due to their composition. Pb-free solder joint defects detected by AXI are shown in Figure 1. One frequently asked question is, What is the difference of the measurement data between Pb-free and SnPb assembly? To determine the answer, we completed an experiment with a Pb-free test vehicle and a SnPb test vehicle using test condition 1 (Pb-free C&A panel) and test condition 2 (SnPb C&A panel).

Experiments and Analysis

The Flextronics test vehicle board (Figure 2) was chosen for this experiment. Board thickness was 1.2 mm with organic solder preservative (OSP) finish. We collected measurements for 12 CSP components, two fine-pitch gullwing and 443 resistors (0402) for measurement comparison (see Table 1 for more information on selected components). The characteristics of Pb-free and SnPb solder are listed in Table 2. The Pb-free solder (TLF-93-206) was from Tamura. All experiments were performed using an Agilent 5DX Series 3 system.

Gage R&R study. One weakness of AXI is the consistency of its results. Defect numbers are not the same when the same board is tested on the same AXI system several times for SnPb assemblies. We wanted to see how the Gage Repeatability & Reproducibility would differ for Pb-free solder joints. We tested one board with three operators, and the tests were repeated three times by each operator. The AXI machine was calibrated with the Pb-free C&A panel first. We chose 30 pins for data analysis on the specific components (BGA, fine pitch gullwing, RES0402) with SPC tool MINITAB. Figure 3 shows Gage R&R results for the BGA diameter from MINITAB. The standard deviation [StdDev(SD)] for a source is the square root of the variance component for that source. Study Var is six times the standard deviation for a source, which means the calculation is based on 6 s. The % Study Var is the % study variation, which estimates how well the measurement system performs with respect to the overall process variation and is independent of the Tolerance. The % Tolerance (SV/Toler) is the Precision/Tolerance (P/T) ratio that is appropriate for evaluating how well the measurement system can perform with respect to the specification, which is dependent on the process tolerance. We use the specification tolerance ±20% for the BGA diameter. The total Gage R&R is 10.84 for the % Tolerance.

We input different specification tolerance numbers (±20 to 65%) into MINITAB for Gage R&R and wanted to have the total Gage R&R number less than 30% (Table 3). We expected that Gage R&R results would be less than 30% with tolerance ± 20%. In Table 3, the reproducibility variations are less than 10% for all joint parameter measurements taken. However, the repeatability variation showed good results only for BGA diameter, BGA thickness, fine-pitch gullwing fillet length and resistor pad solder thickness with tolerances less than 30%. The results of the Gage R&R for BGA voiding percentages, fine-pitch gullwing heel thickness and center thickness need improvement. The results are not surprising since all our experiments for SnPb showed that the repeatability of measurements for these particular components was not a strength of AXI. The BGA voids of the Pb-free TV are very small (diameter less than 0.005"). It has not been proven that AXI can detect voids this small in size. For this particular size BGA, AXI can detect voids that do not meet IPC-7095. Therefore, some false calls are expected. We use a transmission x-ray as a complementary tool to verify these small size BGA voids.

Pb-free TV study. The team measured 12 BGAs (2648 pins) from nine Pb-free TVs with TC 1 (Pb-free) and 2 (SnPb). A total of 47,664 data points were taken for BGA diameter measurements, and another 47,664 data points were taken for BGA thickness. Analysis was performed using Mood Median test, with a 95% confidence level. If P <0.05, there is statistical difference; if P$ 0.05, there is no statistical difference. Most BGA diameter measurements appeared different with different test conditions except U10, U20 and U15. U10 and U20 were the same package type with a pitch of 0.5 mm. The remainder of the BGAs had a pitch of 0.4 mm. (Note: U5 [P=0.041] and U15 [P=0.157] have only 36 pins.) By calculating the percentage difference, we concluded that the measurement data collected under TC 1 was almost the same as that collected under TC 2. The average difference observed was -0.62% using Eq. 1. The difference of other parameters is given using this equation. All BGA thickness measurements were statistically different under TC 1 versus 2. The measurement data with Pb-free TC 1 was larger than with SnPb TC 2. The average difference was 5.09% under different test conditions; the smallest was U3 (1.43%), and the largest was U10 (9.76%).

% difference = 100x (average test 1 - average test 2) / average test 2       (Eq. 1)

For fine-pitch gullwing, we tested two components with 256 pins on nine boards. A total of 13,824 data points were analyzed for heel thickness, center thickness and fillet length. The average measurement of heel and center thickness is listed in Tables 4 and 5. The last columns are the grand average and standard deviation, respectively. Using Mood's Median Test, fine-pitch gullwing heel thickness, center thickness and fillet length of the Pb-free TVs were different under TC 1 and 2. The average difference of TC 1 versus TC 2 was 8.89% and 7.35%, respectively, for the heel thickness and center thickness. It is not surprising that the average of difference is -1.46% for the fine-pitch gullwing fillet length because different test conditions should not affect to the fillet length measurement.

For the resistor, 7,974 data points (joints) of 443 RES0402 components were analyzed with MINITAB. The resistor pad thickness was different under the different test conditions. The average difference was 8.81% using equation 1 under different TCs. The results indicated that the resistor pad thickness was larger with Pb-free TC 1.

SnPb TV study. Similar studies were completed with five SnPb TVs. Differences between TC 1 and TC 2 were obtained using Eq. 2. The average difference of the SnPb TV was -0.14% and -7.05%, respectively, for BGA diameter and thickness. The average difference was -8.85%, -6.25% and 1.51% for fine-pitch gullwing heel thickness, center thickness and fillet length. The average difference of resistor pad thickness was -7.49% under TC 1 and 2.

% difference = 100x (average test 2 - average test 1) / average test 1       (Eq. 2)

Summary

AXI is a powerful tool for process characterization during the transition to Pb-free solder. There is less than a 10% difference for both Pb-free and SnPb TVs under different TCs. Based on this study, to get accurate measurement data, each AXI machine should be calibrated with either a Pb-free or a SnPb C&A panel in accordance with the manufacturing process being used. We continue to use AXI measurement data to verify our Pb-free process, especially to check BGA voids to optimize the profile.

Bibliography

• Glen Leinbach, "The Transition to Pb-free: Pb-free Soldering and the 5DX," Agilent Technologies 5DX Users' Conference Proceedings, March 2005.

• Zhen (Jane) Feng, Eduardo Toledo, Jonathan Jian and Murad Kurwa, "Reducing BGA Defects with AXI Inspection," Circuits Assembly, July 2005.

• Zhen (Jane) Feng, Jacob Djaja and Ronald Rocha, "Automated X-ray Inspection: SMT Process Improvement Tool," SMTA International Proceedings, September 2002.

Acknowledgments

Flextronics Youngsville Engineering team, Agilent support team, CSA support and Dage support team. Shaghayegh (Sharon) Rahmani, Richard Botts, Ken Batchelor, Duston Yau, Toan Nguyen, Mark Evans, Jane Phan, Barbara Koczera, Ken Taber, Stig Oresjo, Rajiv Balsavar and Joe Fisher.

Ed.: This article was first presented at IPC Apex in February 2006 and is reprinted here with permission of the authors.

Zhen (Jane) Feng, Ph.D., is senior staff engineer - Americas at Flextronics (flextronics.com); jane.feng@flextronics.com. Eduardo Toledo, Dason Cheung, Jeff Newbrough and Murad Kurwa are with Flextronics.

• Prev
• Next