Issue link: https://iconnect007.uberflip.com/i/304995
May 2014 • SMT Magazine 25 HEaD-IN-PILLoW X-Ray INSPECTIoN continues to detect HIP effectively. Figure 7 shows AXI 4 measurement solder area at pad level for detected pins, escaped HIP pins, and false call pins for board #495; and all pins for #266. At site 3, with the settings listed in Table 7, AXI 4 detected two HIP pins (no es- caped) for board #266 without any false call; and detected 43 HIPs (escaped 2 HIP) with one false call for board #495. The two escaped and one false call images are listed on Figure 8. As other AXI machine, there is still no very clear difference between good and detective HIP pins as shown in Figure 8. Table 8 listed the test summary from each site with AXI 4 which has better performance than AXI 1 for both detection percentage and false call. Comparison and Improvement All four AXI machines have capability to de- tect BGA HIP defect with different success lev- els. The AXI performance is based on program algorithms, its thresholds' setting, and machine testing conditions. We also collected data for gage R&R from these AXI machines from the different sites. The gage R&R for BGA ball di- ameters are less than 30% for all AXI machines. The gage R&R for middle ball diameters are less than 10% for some AXI machines. However the gage R&R for open outlier, solder area, neighbor outlier are not what we expected because most of them are at the boundary level: around 30%. This is why AXI testing results are not repeat- able. For example, please refer to the AXI 1 test summary shown in Table 9. There are nine test cycle results with board #266 and #499 at AXI 1 machine in our manufacturing (Table 9). It is easy to see the difference of AXI 1 test results with the same algorithm thresholds settings (pad < -3 as defect; mid ball > 2 as defect). Current AXI technology still has plenty of room for improvement in detecting HIP due to its hardware and software limitations. So, what can we do in order to optimize the AXI HIP detection capabilities? At Flextronics, we share our best practices from site to site; as with this project, we suggest setting these algorithm thresholds settings (pad < -3 as defect; middle feATuRe Table 7: Algorithm thresholds with AXI 4. Figure 7: axi 4 "solder area" measurement data for pad slice from site 3 (boards #495 & #266).