Chwee Liong Tee Hoon Chiow Koay Intel Corporation: Kulim, Malaysia The deadline for converting manufacturing lines to lead free processes is getting closer each day. However from a test perspective, are we ready to meet the challenges created by lead free? This paper highlights the challenges faced and makes observations and recommendations for testing lead free PCBAs using an Automated X-Ray Inspection (AXI) machine. Figure 1 shows that an image captured using transmissive x-ray is not able to isolate solder joints on double-sided boards. Figure 2 shows that 3D x- ray is able to isolate solder joints on either side of the board. This capability enables operators in the manufacturing line to diagnose manufacturing faults effectively. On July 1, 2006, European Union Legislation restricting the use of lead in electrical and electronic equipment will go into effect. PCBA manufacturers are scrambling to meet this deadline by converting their existing tin-lead process to lead free, a change that poses serious challenges to in-circuit test (ICT), automated x-ray inspection (AXI), and automated optical inspection (AOI) test methods. This paper focuses on the impact of this change on 3D AXI machines for several reasons. First, they are widely used in the PCBA manufacturing industry due to board design trends such as reduced test-point access that make ICT less viable. Second, some OEMs have instructed their contract manufacturers to add AXI to their production lines, since it can be used to inspect ball grid array (BGA) components, which cannot be visually inspected. Third, advances in computing power have enabled faster image processing thereby reducing AXI test times. And fourth, the 3D AXI is preferred over the 2D transmissive x-ray inspection because of its ability to isolate faults at the top or bottom of the PCBA. Figure 1: 2D Transmissive compared to 3D X-ray images (Source Agilent Technologies Inc.) Although the 3D AXI is a promising tool for inspecting lead free PCBAs, its challenges for the test programmer are becoming apparent. This paper will highlight some of those challenges so they can be met in a timely manner by both factories and AXI manufacturers. According to AXI suppliers, the machine is capable of testing lead free PCBA, because solder defects are captured by measuring the shape of the solder joints. Since these are very similar for tin-lead and lead-free joints, AXI will work with lead free boards. Figures 2 and 3 below show x-ray images of tin lead solder and lead free solder respectively, both of which provide enough contrast for good solder joint images. Page 1 of 6
The 3D AXI analyzes images using 256 gray scale values which are correlated with known solder thicknesses. These thickness-to-gray scale values are stored in a reference table which is then used, along with defect algorithms and acceptance thresholds, to assess the quality of solder joints under inspection. Figure 2: X-ray images of tin-lead solder (Source: Agilent Technologies Inc.) The programmer sets acceptance thresholds for solder thickness at critical regions of interest (ROI) for a particular type of solder joint. During inspection, defects such as open joints or insufficient solder are identified when the gray values at the ROI do not meet the required thresholds.! A design of experiment (DOE) was conducted to study the impact of lead free on AXI methodology. Objectives include the following: Figure 3: X-ray images of lead free solder (Source: Agilent Technologies Inc.) To study the capability of AXI for inspecting lead free solder joints To study AXI yield, false calls, and escapes when inspecting lead free solder joints To measure the impact of lead free solder inspection on AXI test time To study the impact of lead free on BGA joints To compare tin-lead and lead-free solder joint dimensions 3D AXI is used for this DOE. The boards used are of the mini PCI form factor. There are more than 1000 joints per board, and there are two BGA ICs on board. Current manufacturing process does not have ICT because there are no testpoints on the board. Functional test is the last test after AXI. Page 2 of 6
"# AXI capability for inspecting lead free solder joints %# $ This DOE studied three types of solder joints present on PCBAs: gullwings, BGAs, and 0402 resistors. Tin-lead and lead-free processes were used to understand the difference in measurements between good and open solder joints for all three types of solder joints. A bare reflow board was used to establish a baseline for open solder joints. A sample size measurement of 32 gullwing, 588 BGA, and 904 resistor joints were collected in this study. Results showed that for lead free solder joints, the measurements used to distinguish between good and open joints were smaller than those for tin-lead joints, particularly for gullwing and BGA solder joints. Gullwing joints: For gullwing joints, the regions of interest are placed at the heel, the center, and the toe of the solder joint as shown in Figure 4. The AXI open signal threshold is used to capture gullwing open joints by calculating the difference between the solder thickness at the heel and that at the center of the joint. A good joint, as shown in Figure 5, will show a positive value for open signal, because the solder thickness at the heel will be greater than at the center. Conversely, an open joint will show a negative value for open signal, because the center region of the gullwing is then typically higher than at the heel. The greater the difference in good versus open joints, the easier it is for the programmer to set a threshold that will clearly distinguish between good and open joints. Figure 4: For gullwing joints, open signal is the difference between solder thickness at the heel (blue) and center regions (yellow). Table 1 and Figure 5 show open signal results obtained for gullwing joints. They show that tin lead solder will produce a clear distinction between good and open joints, while lead free solder only produces a medium separation. Thus it is easier for programmers to set a threshold that can more effectively isolate good and open joints for the tin lead process than for the lead free process. Programs for lead-free inspection can therefore be expected to result in a higher number of false calls if the programmer sets a tighter threshold and escapes if a looser threshold is set. False calls occur when follow-up inspection finds no actual defect associated with the indictment; while escapes occur when the test fails to indict real failures. Table 1: Larger separation between good joints and open joints for tin lead process compared to lead free process #& '& Figure 5: Lead free gullwing joints show smaller separation between good and open joints. Page 3 of 6
BGA joints: For BGA joints, a maximum of five slices can be set. However, the programmer will normally set only three slices one at the center of the BGA (slice 1), one near the package level pad (slice 2) and one at the board level pad (slice 3) as shown in Figure 6. Since a bare reflow board is used as a baseline, the AXI s BGA insufficient algorithm is used. The insufficient algorithm allows the test programmer to set threshold for BGA ball diameter or BGA solder thickness which can effectively detect missing BGAs on the bare reflow board. Figure 7: Lead free BGA joints show bad separation between good and open joints. As with gullwing joints, it is easier for programmers to set a threshold that can more effectively isolate good and open joints for the tin lead process than for the lead free process. In Figure 7, we see that good and open BGA joints actually overlap each other, making it impossible to set a viable threshold. Resistor joints: For resistor joints, the AXI threshold used to differentiate good and open solder joints is called pin edge delta thickness. This measurement represents the difference between solder thickness at the peak fillet (blue region) and that at the pad under the component (yellow region). Figure 6: BGA slice settings Table 2 and Figure 7 show insufficient results obtained for BGA joints measured at the center slice. Table 2: The tin-lead process produces a larger separation between good and open joints than the lead free process. Figure 8: Pin edge delta thickness is the solder thickness difference between blue region and yellow region. A good solder joint, as the one shown in Figure 8, would have positive value, while an open solder joint would have negative value. The greater the difference, the easier for the programmer to set an optimum threshold that would isolate good and open solder joints without creating excessive false calls or escapes. There were total of 904 resistor joints in this sample. Table 3 and Figure 9 summarize the results obtained for tin lead and lead free processes for the resistor joints. Page 4 of 6
Table 5: No impact on test time for tin lead vs. lead free Impact of lead free on BGA joints Table 3: Lead free and tin lead resistor joints both show medium difference in separation between good and open joints. There are 2 BGA ICs on the board used in this study, IC #1 with a pitch of 0.5mm and IC #2 with a pitch of 0.5 mm. One of the noticeable affects of the lead free process on BGA joints was an increase in solder voids, which appear as lighter images than solid balls in Figure 10. However, the voids percentage was still within specification limits, as shown in Table 6. ' '( Figure 9: Pin edge delta thickness is the solder thickness difference between good and open solder joints for tin lead and lead free process. The results demonstrate in the case of resistor joints that the introduction of the lead free process does not have significant impact on the AXI. AXI yield, false calls, and escapes when inspecting lead free solder joints Table 6: Void percentage is higher for lead-free than for tinlead process. Manufacturing is setting a yield goal of 95% at the AXI station. With the tin lead process, this target was achievable. However, with the lead free process, the yield was only 92.3% as shown in Table 4 below. Results also showed that false calls for the lead free process tripled due to voids formation on all types of joints, but especially on BGAs. Figure 10: More voids in lead free than tin-lead BGA joints. Table 4: Lead free process caused lower yield than tin-lead process at the AXI station. Test Time As shown in Table 5, the lead free process has no significant impact on AXI test time. Solder dimensions study Average solder thickness of BGA balls, gullwing joints, and 0402 resistors and capacitors was studied. As shown in Table 7, the dimensions did not differ significantly from the tin lead process. Page 5 of 6
In conclusion, the lead free process poses great challenges to both test programmers and AXI manufacturers: The AXI programmers must learn to fine tune programs so that they work as well as they do for the tin lead process. AXI suppliers need to improve their software to work more robustly in a lead free environment. " Table 7: Solder thickness comparisons between lead free and tin lead process )" The lead free process caused an increase in voids for all solder joint types, but especially in BGA joints. Since lead-free images appear lighter than tin-lead images, the problem is to adjust and refine the AXI software so that it can distinguish grey scale variations within the lead-free range to identify and fail genuine voids 1. Stig Oresjo and Robert Ling: The importance of Test and Inspection when implementing lead-free manufacturing, Agilent Technologies, Inc. 2. Jon O Conell, Senior Consultant Agilent Technologies UK Ltd: Automated Test group, Study and recommendations into using lead free printed circuit board finishes at manufacturing in circuit test stage. 3. Stephen F. Scheiber:, X-ray s 2-D vs. 3-D debate: To slice or not to slice? (2001) $# The DOE concurs with the supplier that AXI is capable of detecting open joint defects for the lead free process. However, the increase in false calls shows that AXI is currently less efficient for lead free than it is for the tin lead process. The formation of many voids in lead free solder joints, particularly for BGAs, has caused the number of false calls to triple. Even though the percentage of voids detected remained within specification, this increase will be something to watch carefully as lead-free processes roll out. The lead free process also resulted in lower solder thickness for gullwing joint types compared to the tin lead process. However it must be noted that the introduction of lead free process did not contribute significantly to AXI test time, yield, or escape defects per million (dpm). Page 6 of 6