INTRODUCTION
The semiconductor company in this case study is a leading international provider
of IC packages that powers the products consumers use every day. The company
was facing one of its greatest challenges. Defects have occurred in daily production
and affect the daily production of IC packages in end of line as shown in Fig.
1. Some of the types of defects are voids, cracks, incomplete mold, fail
stand-off height, broken package and marking defects. Appropriate actions are
needed to identify root causes, workout and execute solutions. The management
decided to confine the project on the end of line process of micro department.
The lessons learnt from this pilot project will be implemented in other projects
involving other departments. For effective problem analysis, it is important
to follow a logical approach using specific tools to arrive at root causes.
At each stage, a tool is chosen for the appropriate circumstances to generate
answers. These answers may then generate other questions or clues on the problem
analysis trail and other sets of answers are found. This investigative process
continues until the root causes are finally determined. It should be noted that
choosing the wrong tool at any stage on the problem analysis trail may lead
to dead ends (Kumar and Sosnoski, 2009; Ho,
1993; Gwiazda, 2006; Hagemeyer
et al., 2006). As a simple guide to choosing the appropriate tool,
one should begin to understand the outcome of using any particular problem solving
tool and the information that is available at the time to make the use of the
tool applicable (Bruce, 1990; Juran,
2009). For example the Ishikawa Diagram is used to identify many possible
causes for a known effect or problem.
It organizes ideas into helpful categories that are used to identify possible
causes for a problem. Stratification is a technique used in combination with
other data analysis tools. When data from a variety of sources or categories
have been lumped together, the meaning of the data can be impossible to see.
This technique separates the data so that patterns can be seen. The process
map or flow chart is one of the oldest, simplest and most valuable techniques
for depicting and organizing work. It is used to show the sequence of events
to build a product. The pareto analysis is a statistical technique that is used
for the selection of a limited number of tasks which produce significant overall
effect. It employs the 20/80 rule, the idea that by doing 20% of the work one
can generate 80% of the benefit of doing the whole job. Or in terms of problem
analysis, a large majority of problems or defets (80%) are actually caused by
a few vital causes (20%) (Juran, 2009).
|
Fig. 1: |
Process for IC packages in end of line |
Another useful problem analysis tool, the Taguchi method, can be employed as
a mechanism for evaluating and implementing improvements in products, processes,
materials, equipment and facilities. As a result of studying the key parameters
that control a process, improvements can be realised for desired characteristics
that can substantially reduce the number of defects (Klein,
1996; Anthony, 2006; Tong et al.,
1997). The tools mentioned thus far, were used to analyse the problem, identify
possible causes and formulate solutions to effectively reduce the defects in
the IC packages.
MATERIALS AND METHODS
The overall methodology has been summarized in Fig. 2. IC
packages were inspected under 30x scopes after each process and the data of
defects were collected.
Table 1: |
L25 Orthogonal Array |
|
The data for IC packages with the most defects were stratified to highlight
patterns if any in the defects. Next, pareto analysis was carried out to segregate
major contributing defects categories. The Ishikawa diagram was then used to
draw out possible causes for the major defect. To understand the effect of the
combination of contributing parameters for the identified possible cause, the
Taguchi method was then employed.
Table 2: |
Values of parameters set for each experiment |
|
Experiments were conducted with 6 parameters and 5 levels. The appropriate
orthogonal array is L25 as shown in Table 1.
For the experiment, the following factors are assumed:
• |
Other variables besides mold parameters are assumed constant |
• |
The experiments are carried out on a fixed auto mold machine |
• |
The surrounding condition is assumed to be constant |
• |
The same operator was assigned to operate the machine |
• |
The experiment is conducted on dummy frame (without die) |
The conditions set for each experiment are listed in Table 2.
RESULTS AND DISCUSSION
The result from defects inspection of all the packages is summarized in Table
3. Three IC packages, 6SOT23M, 5SC70M and 6SC70M were identified to have
the highest number of defects. These 3 packages in fact make up a substantial
70% of the company output from Micro End-of-Line and therefore deserves more
focus. To facilitate a closer inspection of the 3 packages, stratification was
employed. Figuer 3-5 summarize the stratification
exercise.
Table 3: |
Defects according to package type |
|
From the 3 stratification charts above, the defects were categorized into process
specified defects. As shown in Table 4, 5 and
6, there are only 5 manufacturing processes in micro end of
line. These are molding, laser mark, high pressure water jet, plating and trim,
form and singulation. The others are inspection and packing processes. Defects
such as incomplete mold, void, fail vertical offset, flashes, compound leaking,
flake surface and unclean package are categorized under molding processes defects.
Lead width, fail stand-off height, bent lead, missing lead and no forming are
categorized under trim, form and singulation process defects. Marking defect
comes under its own category and originates from the laser marking process.
While chip, dented lead, crack and broken package is categorized under other
group because these defects can be caused by various processes or combination
of a few processes.
|
Fig. 3: |
Stratification of defects for 6SOT23M |
|
Fig. 4: |
Stratification of defects for 5SC70M |
|
Fig. 5: |
Stratification of defects for 6SC70M |
|
Fig. 6: |
Pareto diagram of defects for 6SOT23M |
Table 4: |
Pareto table of defects for 6SOT23M |
|
|
Fig. 7: |
Pareto diagram of defects for 5SC70M |
Table 5: |
Pareto table of defects for 5SC70M |
|
|
Fig. 8: |
Pareto diagram of defects for 6SC70M |
Table 6: |
Pareto table of defects for 6SC70M |
|
Table 7: |
Consequences when mold parameter out of control |
|
Using the Pareto analysis, it was also observed that molding is the process
that contributes most to the defects of the 3 types of packages which cover
62.96% of 6SOT23M, 57.14% of 5SC70M and 73.91% of 6SC70M as shown in Fig.
6, 7 and 8. Therefore, the company then
decided to look at the molding process in more detail as out of control mold
parameters produce consequences as shown in Table 7.
Ishikawa analysis of incomplete mold of 6SOT23M drew out possible causes. These were verified as shown in Table 8. To optimize the mold parameters the Taguchi method was employed. The results of the experiment are shown in Table 9.
It can be seen that experiment No. 15 has the lowest defect number yield. This mold parameters configuration was suggested to be used instead of the current mold parameters because it yields the least number of defects. Shown below is the sample calculation and tabulation of the SN ratio.
Table 8: |
Verification of possible causes for incomplete mold of 6SOT23M |
|
Table 9: |
Data collected from experiments |
|
Table 10: |
SN ratio values |
|
The other values of SN ratio are calculated as above and tabulated in the Table
10. The response table to calculate an average SN value for each factor
is shown in Table 11.
A sample calculation for Factor B (Clamp Pressure) is shown below:
Table 11: |
Average SN values |
|
Table 12: |
Values of SN and ranking |
|
The value of SN for each factor and value of the effect of the factor is shown in Table 12. It can be seen that clamp pressure has the largest effect on the defect number and transfer pressure has the smallest effect on defect number.
The effect of this factor is then calculated by determining the range.
CONCLUSION
The processes in Micro End of Line produced a certain amount of defects. The problem analysis identified all the defects and categorized them into different categories as well as reported the identified possible causes. The investigation zoomed into the molding process because the evidence showed this process was contributing the most defects. The use of the Ishikawa Diagram and the Taguchi method successfully identified the more critical parameters as well as the best parameter configuration that produced the least number of defects. It can therefore be concluded that the mentioned problem analysis tools can be effectively used to reduce the number of defects in IC packages provided they are employed appropriately.