New Process Optimization Example: Target Value Manipulations And Variability Reduction Doe
A good example of new process optimization is the introduction of fine pitch SMT into the manufacturing process. Fine pitch SMT requires smaller solder paste deposition of solder bricks with a target height of 0.005’’, and the quality is enhanced by the variability reduction of ^ process. The fine pitch SMT project is a succession of small DoEs that leads to achieving the new product quality target It can be summarized as follows, with the data information listed in Table 8.16 for the aver, age and Figure 8.9 for variability of the SMT processing parameters;
1. The quality characteristics were defined as achieving a solder paste height in the solder deposition process with a target of 0.005tt, with minimum variability.
2. The quality characteristics were measured on a test PCB containing many of the fine pitch components used.
3. Solder paste thickness was the average of four measurements in each PCB, measured at the corners of specific components. The corner represents the most difficult location in which to achieve uniformity.
4. The measurements were repeated on two PCBs, for determining variability. They were expressed as S/N for the smaller-is-better case, which is the same as -10 log variance. This S/N level was used instead of the S/N nominal formula since there were two separate analyses, one for average and the other for variability.
5. A full factorial L8 orthogonal array DoE was initially used to select the material supply for the process. Factors included the selection of the paste, stencil thickness, and the squeegee hardness.
6. For the processing methods selection, an L9 orthogonal array was used in saturation design, with four factors at three levels including squeegee speed, pressure, down-stop, and snap off distance.
The same experiment was used to analyze average and variability data.
7. The stencil was wiped off between successive prints on the PCBs.
An automatic height laser machine recorded the measurements.
8. Average and variability analyses were calculated for the experiments as shown in Table 8.16 and Figure 8.9. Some of the data indications are:
• The S/N for variability reflect mostly negative numbers due to the -10 log formula for variability conversion. The desired outcome for each factor is the level with the most positive value in all cases of variability analysis.
• Levels 2 and 3 of Factor B (squeegee down-stop) have the same average but different variability effects.
• Experiment line 8 had the least variability, but not at the target value (0.005).
• One additional row in the ANOVA analysis, not shown in the previous Chapter 7, is the error due to replication. It is calculated from the subtraction of the total SSp of the factors from the SS of the total.
SSError = SStotal - SSA – SSB - SSC - SSD
• Factor A (squeegee speed) was not significant and was pooled into the error.
• The S/N graphical analysis data in Figure 8.9 was obtained from
analyzing the experiment S/N data versus each experimental i line, similar to the average analysis.
• Levels selected to reach the target value of 0.005 include those levels with the lowest variability. For example, B2 was selected instead of S3 because it was more positive in the S/N calculations, even if both scored the same value for the average analysis.
• Subsequent DoEs are needed to continue to account for the 44% error of the experiment. Interactions of the factors should be studied as well as more repetitions of the experiments and more factors and levels with lowest variability.