-
- News
- Books
Featured Books
- smt007 Magazine
Latest Issues
Current IssueComing to Terms With AI
In this issue, we examine the profound effect artificial intelligence and machine learning are having on manufacturing and business processes. We follow technology, innovation, and money as automation becomes the new key indicator of growth in our industry.
Box Build
One trend is to add box build and final assembly to your product offering. In this issue, we explore the opportunities and risks of adding system assembly to your service portfolio.
IPC APEX EXPO 2024 Pre-show
This month’s issue devotes its pages to a comprehensive preview of the IPC APEX EXPO 2024 event. Whether your role is technical or business, if you're new-to-the-industry or seasoned veteran, you'll find value throughout this program.
- Articles
- Columns
Search Console
- Links
- Events
||| MENU - smt007 Magazine
Detect Defects with SPI & AXI
December 31, 1969 |Estimated reading time: 12 minutes
Solder paste inspection (SPI) systems can detect the majority of paste-related post-reflow defects upstream, before they occur. Adding 3-D SPI after screening can reduce reliance on AXI. This article details a study conducted with SPI, AXI, and transmissive X-ray (2DX) that investigates the correlation between solder paste anomalies found BEFORE reflow, and resulting solder defects found after reflow.
By Zhen (Jane) Feng, Ph.D.; Alex Garcia; Scott Kingery; Murad Kurwa; Walid Meliane; and Thomas Munnerlyn - Flextronics International
As PCB assemblies become more complex, SPI and automated X-ray inspection (AXI) systems are widely used. The use of SPI to inspect/measure solder paste height, area, and volume has been considered to reduce reliance on AXI due to beat-rate concerns. When compared to other machines on the SMT line, AXI Takt time is longer. Several years ago, a paste inspection study was conducted with solder paste inspection (SPI) and AXI systems. For 1,677 boards, more than 2,000 paste deposits had below 65% normal paste volume, none of which was detected as a “defect” after reflow. On the other hand, AXI found and confirmed 46 defects after reflow, none of which were found by the paste inspection system.1
Figure 1. Lead-free test vehicle.
Now, more-capable 3-D volumetric SPI systems are available in the industry. This article documents a study that answers the questions: What is the correlation between SPI and AXI using both attribute and variable data? How should correct thresholds be set for SPI to catch defects with a low false-call rate? What kind of products benefit from reduced AXI test coverage by adding capable SPI? This article investigates the correlation between SPI and AXI with attribute and variable data, and details a two-phase experiment.
Phase One
Evaluate SPI system capabilities, including Cp and Cpk; gage repeatability and reproducibility (gage R&R); and orientation repeatability tests with 0°, 90°, 180°, and 270° board loadings.
Phase Two
Find a correlation comparing images and variable data from SPI, AXI, and 2DX, and look for the correlation between paste defects found just after solder paste printing, and compare them to real solder joint defects found after reflow. A lead-free qualification vehicle* was chosen (Figure 1), and measurement data was taken from the SMT line. Data was also collected from a customer product using tin/lead.
Methodology
The experiment used an SPI system**, an AXI system***, and a transmissive X-ray system****. For phase one, SPI Cp and Cpk, gage R&R, and orientation repeatability were studied. A standard board was used for Cp and Cpk, and the lead-free test vehicle was used for gage R&R and orientation test experiments. This vehicle is a two-layer board with X/Y/Z size of 152 × 216 × 1.57 mm. SAC (95.5Sn/3.8Ag/0.7Cu) solder was used with a soak (oven temperature of 150°-217°C) time between 120-125 seconds. BGA and QFP peak temperature was 236.7°C. Components used are listed in Table 1.
Cp and Cpk. The standard board has 2.97-mm round pads with 0.1475-mm height. The board was inspected 50 times at 0° orientation, was rotated 90° degrees, and then inspected another 50 times for solder height, area, and volume.
Figure 2. Orientation test results.
Gage R&R. One lead-free board was tested for SPI gage R&R. Three operators tested the board three times each, for a total of nine measurements. The SPI original data file has test-pin IDs only, and was converted to corresponding component and pin numbers. The same 30 pins were selected for volume, area, and height gage R&R from component BGA (U5, pitch=31.5 mils), QFP (U3, pitch=19.7 mils), and resistor (RES 0402).
Orientation test. One lead-free test vehicle was printed and loaded in 0° orientation. The pads were then inspected for volume, area, and height for one BGA (U5, 308 pins), two FP-Gullwing devices (U3, and U4, 208 pins each), and 20 RES 0402 (40 pins). In total, the boards were inspected four times at 0°, 90°, 180°, and 270°.
Lead-free test vehicle. All 30 boards were tested at SPI first, and then sent through pick-and-place and reflow. They then were tested in the AXI system, and the results (images and measurement data) were compared between SPI and AXI. Additional 2DX was used, if the AXI image was not clear, for example small BGA voids. This test helped establish the “SPI threshold setting,” for example, what paste volume measurement will trigger a “no go” at the SPI machine? SPI threshold settings are shown in Table 2.
Figure 3. Solder insufficient images for Pin #14 of U5.
Customer production board. Thirty 10-layer, 620-mil-thick customer production boards using tin/lead solder paste were tested in SPI and AXI. 2DX was also used to identify some images to see how to reduce AXI testing by adding SPI for a stable SMT process.
Lead-free test vehicle with induced defects. A stencil was created with varying apertures to change paste-deposit volumes and create a larger defect spectrum when studying the correlation between the values measured by the SPI system and the defects found by AXI post-reflow. Two surface finishes: Ni/Au and organic solder preservative (OSP) were present. The correlation between SPI, AXI, and 2DX with the lead-free test vehicles were studied with “induced” defects. Two BGAs, five QFPs, one PLCC, and 20 RES 0402 were selected for the experiment. The stencil-aperture area was designed from 20% to 150% of the target number for assigned locations. A total of eight boards were used: two PCBs without components and six PCBAs with components mounted. Both variable and attribute data were collected for the correlation study. However, only the BGAs have the variable data automatically from SPI, AXI, and 2DX. The remaining components have only variable data from SPI and AXI. The SPI threshold settings were adjusted individually for this experiment.
Experiment
Statistical process control (SPC) was used for Cp/Cpk and gage R&R analyses. 2DX attribute results depend on AXI images, including variable data and solder joint quality.
Cp and Cpk. The height lower specific limit (LSL) and upper specific limit (USL) are 142.5 µm and 152.5 µm, respectively. The LSL and USL for area and volume are -4% and +4% of the target number. Target LSL and USL numbers are listed in Table 3. One-hundred data points were analyzed each for height, area, and volume using SPC with six sigma tolerance. Results indicated that SPI has excellent accuracy and repeatability capabilities for height, area, and volume measurements with the standard board.
Gage R&R. The SPC tool was used to calculate gage R&R with six sigma and tolerance of ±20. The gage R&R of FP-Gullwing and RES 0402s are better than BGA, most were excellent (<10%). SPC gage of reproducibility for BGAs was higher than repeatability. The reason was unclear; total gage R&R for BGAs, FP-Gullwings, and RES 0402s were acceptable.
Orientation test. The board was tested four times, and results were averaged. The percentage difference was calculated using the equation: percentage difference = 100% (maximum reading average - minimum reading average)/total average, where maximum average and minimum reading average are from the four test runs. Figure 2 shows percentage difference of orientation test results. All percentage differences for BGAs, FP-Gullwings, and RES 0402s were less than 8%, which is acceptable. Height has better results than area and volume. All percentage differences of RES 0402s for height, area, and volume area are below 3%. Volume and area differences for BGAs and FP-Gullwings are larger than 5%; however, height differences were good.
Lead-free test vehicle. Images and measurement data from SPI, AXI, and 2DX were compared for all 30 lead-free vehicles. There were 14 real defects in total. SPI detected six; AXI and 2DX detected all of them. This result was encouraging because data showed the correlation between SPI and X-ray is 43%. Results are listed in Table 4.
Figure 4. Images of pins #121-124 of U3. (L-R) Q10 board (SPI) Q10 board (2DX).
SPI, AXI, and 2DX have correlations for some solder joint defects. For example, on a U5 BGA, with a pitch of 31.5 mils, SPI, AXI, and 2DX indicated pin #14 of board Q8 had insufficient solder (Figure 3). All three SPI, AXI, and 2DX machines indicated pin #14 had lower measurement data. The data show a strong correlation for pin #14. When considering chip-scale package (CSP) U43 with a 20-mil. pitch size, SPI caught five of eight defects. Three boards had an insufficients on pin #2, and all machines caught them.
Figure 5. SPI images for pins 101-102 of Q10.
SPI, AXI, and 2DX do not have good correlation for some solder short defects. SPI caught four pins (#121-124) of U3 on board Q10 as clear solder bridging (Figure 4); however, the short disappeared as a result of wicking in the reflow process. For pins 101-102, the short was induced by the placement process after SPI inspection. Figure 5 shows SPI images of no solder bridging after solder paste, but further AXI and 2DX images post-reflow show solder bridging. Additional data is needed to indicate how large the bridge will be after reflow with different components. SPI cannot detect defects induced after solder paste printing, such as misalignments and BGA voids. For very-small-pitch CSP U43, SPI does not have a strong correlation. This means that not all insufficient solder defects for CSP after reflow will correlate with lower SPI measurement data.
It was concluded that SPI volume is the best parameter to predict and catch insufficient defects before they occur. All pins with volume <40% results were considered real defects, and pins with volume of <50% depended on area-percentage measurements.
For the 30 production boards, SPI caught 11 defects and AXI found one. These 12 pins were not perfect; however, all pins met IPC standards. For example, AXI called void as 22.2% and 2DX identified the void as <20%.
Eight lead-free test vehicles with induced defect, attribute, and available data were compared for SPI, AXI, and 2DX. Solder volume, area, and height data was collected from SPI for all components; solder area data was collected for BGAs, as was volume data for QFP and PLCC, height data for resistor from AXI, and solder area data for BGA from 2DX. There was a strong correlation (≥0.7) for SPI volume vs. SPI area, AXI area, 2DX area; SPI area vs. AXI area, 2DX area; and AXI area vs. 2DX area for the BGA U1 of the two unpopulated boards. In this study, the correlation between SPI volume vs. AXI area and SPI volume vs. 2DX area for BGAs; SPI volume vs. AXI volume for QFP, PLCC; and SPI volume vs. AXI height for resistor was determined. Most components had a strong correlation except for two: a QFP with 208 pins and one QFP with eight pins. This is because all pins of U46 have the same stencil aperture; it is difficult to have a strong correlation due to small measurement variations compared to the systems’ measurement resolution.
Six populated boards were tested - three had Ni/Au surface finishes and three had OSP finishes. The correlations for SPI, AXI, and 2DX were strong with coefficient C ≥0.7; medium correlation with coefficient 0.7<C/≥ 0.5; and weak correlation with coefficient C<0.5. Resistors have strong correlations for all six boards. QFP U53 and PLCC U54 had strong correlations, except for board J3. IC U46, which had weak correlations for three populated boards because all stencil apertures have the same design, and no defects were created on these locations. Components U3, U4, and U20 were different packages with the same pitch size (19.7 mils), but a different number of pins. Their correlation may be different because U3 and U4 have more pins than U20, and there are different stencil aperture percentages.
To explain why components with more pins have weaker correlations when there is little solder difference, look at correlations from BGA U5 measurement data on the regular lead-free test vehicle. All SPI, AXI, and 2DX indicated the lowest reading at pin #14, and it is an insufficient solder (see authors for figure). However, the correlation coefficient is only 0.290 for SPI and AXI with 26 data points. This is a weak correlation, but the machines have the capabilities to find a difference between good and bad solder joints. The correlation coefficient is small if all pins have the same stencil aperture, such as QFP U46. Available data show that SPI, AXI, and 2DX have a strong correlation for the solder with a large difference, which can be used to set “go/no-go” thresholds.
Attribute data from SPI and AXI also was compared for these six populated boards. The boards were inspected first by the SPI system, and then by the AXI system after reflow. More than 89% of BGA pins had good correlations. Ten percent of BGA pins failed at SPI, but passed at AXI because their solder area was outside the SPI threshold. However, the AXI pass occurred because AXI area data are smaller than 120%, and larger than 85%.
Conclusion
SPI, AXI, and 2DX have good correlation; variable data is strongest for BGA, PLCC, and resistor packages. Setting the right thresholds at which SPI fails a deposit is the key to optimizing this technology on the SMT line. SPI is effective at detecting solder joint defects before they occur - pre-reflow. It also is a good process monitor.
AXI can detect more defect types, such as voids that occur during reflow, and should be part of a process monitoring or audit process. It can detect 90% of defects for assemblies, and is a good SMT process improvement tool with real-time data feedback.2 2DX complements AXI in production to recheck “open” or “void” BGA joints, and can provide clearer images for some defect pins at the boundary of pass/fail. By adding SPI, AXI test coverage can be reduced for stable manufacturing process products, resulting in cost savings. AXI and 2DX are the keys to new product and new package introductions.
* Flextronics lead-free qualification vehicle.** Orobotech Symbion P36 system.*** Agilent Laminography 5DX system.**** Dage 7500 trasmissive X-ray.
ACKNOWLEDGMENTS
The authors thank the Flextronics engineering and production teams in Dallas, Texas; as well as the Orbotech, Agilent, and Dage support teams. The authors also thank the following for their assistance: Dennis Dugan, Jeff McGee, George Willimans, John Upham, Allyn Sedgwick, Michael Xie, Georgie Thein, David Lin, Martin Murguia, King Lee, Dave Dunne, Kim Hyland, Manny Deluna, Eric Bert, Mart Genender, Calvin Griffin, Barbara Koczera, Heidi Hanson, and Joe Fisher.
References
- S. Oresjo and V. Chatrath, “Paste Inspection Study,” Proceedings from IPC APEX, Feb. 2002.
- Zhen (Jane) Feng, Jacob Djaja, and Ronald Rocha, “Automated X-ray Inspection: SMT Process Improvement Tool,” Proceeedings from SMTAI, September 2002.
For a complete list of references and figures, please contact the authors.
Zhen (Jane) Feng, Ph.D., senior staff engineer - Americas, Flextronics, may be contacted at (408) 576-7683; e-mail: jane.feng@flextronics.com. Alex Garcia, process engineer, Flextronics, may be contacted at (469) 229-2150; e-mail: alex.garcia@flextronics.com. Scott Kingery, director of engineering, Flextronics, may be contacted at (469) 229-2379; e-mail: scott.kingery@flextronics.com. Murad Kurwa, vice president of engineering - Americas, Flextronics, may be contacted at (408) 576-7087; e-mail: murad.kurwa@flextronics.com. Walid Meliane, senior manager, process engineering, Flextronics, may be contacted at (469) 229-2405; e-mail: walide.meliane@flextronics.com. Thomas Munnerlyn, senior test engineer, Flextronics, may be contacted at (469) 229-2195; e-mail thomas.munnerlyn@flextronics.com.