Issue link: https://iconnect007.uberflip.com/i/721932
September 2016 • SMT Magazine 41 dix A. This table summarizes the data for the eight test vehicles assembled using the same as- sembly process, as indicated in Table 3 above. A similar table was created for all of the test ve- hicles. (The complete data sets in digital format can be made available to investigators upon re- quest.) Three measurements taken on the 97% tin calibration standard are shown in the last three rows of Appendix A. All thicknesses are given in micro inches. All compositions are giv- en in weight percent tin (the remainder is lead). All tin measurements in excess of 97% are shad- ed pink, and all thickness measurements less than 10 micro inches are shaded blue. The loca- tions for each measurement are indicated by ab- breviations as follows: "E" – Where the lead is protruding from the package "M" – The midpoint of the lead "P" – At the pad "TL" – Top left side of the package "BL" – Bottom left side of the package "TP" – Tab where it meets the pad "TE" – Tab where it is protruding from the package "EL", "ML", and "PL" – For the lead, where it is protruding from the package, at its mid-point, and at the tab, respectively The data was reviewed for consistency and to determine the validity. Some measurements, particularly ones taken near where the lead pro- trudes out from the package (E) of certain com- ponents, were found to exhibit unreasonably low coating thickness readings. Inspection of some of the suspect components revealed that there was bare copper exposed at this location, resulting in the collection of no meaningful data on solder coverage at these locations. Analysis of results Preliminary data analysis has commenced to help with data summary. To begin with, the data was examined for suspect data readings, in light of the issue with unreasonably thin read- ings discussed above. For this preliminary sum- mary any reading with a thickness measurement less than 15 µin was eliminated from the datas- et. The tin and lead content from these samples were considered suspect and would need addi- tional review to be considered accurate. Some basic parameters were examined for significance. ANOVA analysis with back- wards deletion was used to examine the signif- icance of pad size, board type, and soldering process. Board type (HASL or OSP) was found to be an insignificant explanatory variable (F- value 2.34, Pr>F: 12.6%). While pad size was found to be significant during the analysis (F- value 5.8, Pr>5: 1.6%), the difference between means for the two sizes was less than 3%, so on a practical basis it is not expected to be a factor. The differences between soldering pro - cesses is significant, but will require more in depth analysis to determine the specific differ- ences in the processes driving the differences in performance. When rating the relative performance of the different packages and soldering processes, we were more interested in the extremes than in the average behavior. For example, we were con- cerned whether 99% of parts would be compli- ant in a production setting. Therefore, we per- formed our rating based on tolerance intervals. Tolerance intervals are confidence intervals for a covering of a fixed percentage of the popu- lation. For this analysis, we used formulas for one-sided tolerance intervals for a normal dis- tribution. [2] These were calculated for three dif- ference confidence levels and looking at several different percentiles of interest. The data was analyzed at three difference confidence levels: 60% confidence (little con- fidence over expected value), 80% confidence (adding some margin), 90% confidence (high confidence needed for critical systems). The re- sults are summarized in Tables 4–6 that follow. The following color code is used: • Green means the chance of being <97% tin is at least 0.99 • Yellow means the chance of being <97% tin is between 0.9 and 0.99 • Orange means the chance of being <97% is between 0.75 and 0.9 • Pink means the chance of being <97% is between 0.5 and 0.75 • Red means the chance of being <97% is less than 0.5 MITIGATION OF PURE TIN RISK BY TIN-LEAD SMT REFLOW