22nd ASEMEP TECHNICAL SYMPOSIUM

1
A Systematic Approach in Optimizing Critical Processes of High Density and High
Complexity New Scalable Device in MAT29 Risk Production
Using State-of-the-Art Platforms

Antonio R. Sumagpang Jr.
Arnold Rada


NPI - Assembly Manufacturing Department
ST Microelectronics Calamba, #9 Mountain Drive, Light Industry & Science Park II
Barangay La Mesa, Calamba City, Laguna, Philippines 4027

antonio.sumagpang@st.com
arnold.rada@st.com


ABSTRACT

This paper will discuss how to deal with a new product
trend that will survive during production ramp-up. The
project was intended to determine the required process flow
and platforms for high-density and high-complexity scalable
device. Critical processes were also shown and how top
reject contributor was addressed through methodological
way by using Statistical tools and in depth engineering
analysis.
The Scalable Device is one of the newest and latest
developed device in ST Microlectronics Calamba. The
device is considered High Density as single wafer (6) is
equivalent to 400,000 units compared to conventional
device consisting of only 1,000 units. Likewise, it is
considered as High Complexity as State-of-the-Art
platforms are needed to satisfy its output process. Scalable
Device, not to mention that this is the thinnest die (90um)
and with the smallest total package dimension (0.3x0.6x0.28
mm) being process in Calamba, carries a Step Cutting
method of wafers, Compression Molding, and In-Strip
Testing, unlikely to be found on other Semiconductor
Industries.
These state-of-the-art technologies were showcased in this
paper. Compared to the conventional and universal
approach, complex errors and top reject contributor of
identified critical processes were corrected and required
process capability index was achieved.


1.0 INTRODUCTION

In order to survive the past-paced technology in
Semiconductor Industry, one should have a very good
impression from the customer be it internal or external. This
is one of the biggest challenges for any semiconductor
company in order to maintain its competitive market
position and value. Satisfaction is the right word and the
key factor in building good relationship with the customer.
On the other hand failures to provide customer expectation
in terms of on-time delivery will result to possible business
failure.

This critical scenario should be avoided thats why a risk
production (MAT29) is being done in preparation to Full
production mode (MAT30). A total of 10 to 30 lots are line
stressed to capture all hindrances in the production line and
thus corrected immediately to prevent delivery issues.

Misdeliveries or delinquency in view of customer was the
scenario encountered during the ramp-up/line stressing of
new scalable device. With the latest technology trend and
state of the art platforms this paper will discuss how this
burden was turned into milestones when top yield detractors
of critical processes were addressed by in depth engineering
analysis and utilizing statistical tools at early stage of
production.

1.1 The New Scalable Device

The New Scalable Device is a diode w/ single wiring
connection which main function is ESD protections.
Shown below is the package illustration of Scalable device:


Figure 1: New Scalable Device package construction

1.2 Comparison to the size of grain

The New Scalable Device was compared into the size of
grain to illustrate the complexity of its process and device
itself with primary consideration on its total package
dimension. As the size becomes smaller, this becomes more
complex.




Figure 2: New Scalable Device comparison to size of a grain
0.3mm
0.6mm
0.083mm
22nd ASEMEP TECHNICAL SYMPOSIUM
2
1.3 The Applications

Two main applications of Scalable Device are for mobile
phones (USB, Display protection) and Computers (DVI,
HDMI, and ESD Protection).

Figure 3: New Scalable Device Package Application

1.4 Design Rules

Single Die Assembly
Max Die Size: 0.20mmX0.17mm after sawing
(0.22mmX0.19mm before sawing)
Max Die Thickness: 100 um (90+/-10um)
Min Bond Pads Openings: 70 um (Al Metal)
Wafer Sawing Street Acceptance: 55 um

1.5 New Scalable Device Full Process Flow:

Below is the complete process flow for Scalable device
starting from Pre-Assembly to Back-end assembly until Test
and Finish/Packing.


Figure 4: Scalable Device Complete Process Flow

1.6 New Scalable Critical Process w/ identified risks:

Three critical processes were identified using risk analysis.
Evaluation was made before the risk build to accelerate
confidence on MAT29.

Furthermore, Potential Risk Analysis was given contingency
plans and created corrective actions.



Table 1: Scalable Device showing three Critical Processes
identified namely, <1>Wafer Sawing, <2>Compression
Mold and <3>In-Strip Testing


1.7 New Scalable Critical Process Defect Contribution


Graph 1: Scalable Device Defect per Process Pie Chart

Graph 1, Pie Graph shows reject contributor on the
identified critical processes. Wafer Saw, Mold and In-Strip
Test experienced deviations/ output abnormalities as a result
of unoptimized parameters which are normally attributed to
newly introduced device.

1.8 Pareto of Rejects per Critical Process

1.8.1 WAFER SAW Pareto of Rejects (Critical Process #1)

Of the 23% defects from wafer saw process, die chipping
was the top contributor, followed by dice off and broken
wafer.




22nd ASEMEP TECHNICAL SYMPOSIUM
3

Graph 2: Wafer Saw Process Pareto of Reject showing
Chippings as the top contributor. (Actual DPPM values
intentionally not shown)

1.8.2 MOLD Pareto of Rejects (Critical Process #2)


Graph 3: Mold Process Pareto of Reject showing Voids as
the top contributor. (Actual DPPM values intentionally not
shown)

1.8.3 FT (In-Strip Test) Pareto of Rejects (Critical Process
#3)

Graph 4: FT Process Pareto of Reject showing AA (Auto
Align) Fail as the top contributor. (Actual DPPM values
intentionally not shown)
Above Graphs (2, 3 &4) show the top rejects of mentioned
critical processes affecting severely the line stressing mode.
Parameter optimization is one of the factors to be checked as
this type of device is to be built first time in Calamba.
Benchmarking for similar device to other sites is being
considered to have a baselining on critical process
parameters.

1.9 Problem Statement

Top rejects (based on Pareto) of identified three critical
processes substantially affects the yield and delivery during
production stressing performance. Thus, optimization is
highly regarded before it reaches the MAT30 maturity or
full production release.

Further analyses and investigations of failures are made by
collecting actual reject samples from critical processes. This
will serve as lead in the investigations and formulation of
corrective actions.


Table 2: Top Defect Signature of Critical Processes


2.0 REVIEW OF RELATED LITERATURE

We had mentioned the critical processes of scalable device
and their corresponding top reject mode. Below will explain
the details of each processes and how top rejects were
addressed.

2.1 Wafer Saw (Critical Process #1)

As Scalpack is considered as Low K Wafer (very thin),
Sawing becomes a critical process. When abrasive blades
cut or groove material, they are actually grinding and
removing it. The mechanism is similar to that of a metal
saw: the gaps between the teeth of the saw whisk material
22nd ASEMEP TECHNICAL SYMPOSIUM
4
away from the point of processing. These gaps, called chip
pockets, are circled in red below.



Figure 5: Wafer Sawing mechanism how Chipping pockets
occurs.
As shown in the figure 5, new blade has diamonds covered
wholly by the bonding material and no diamonds (hammers)
are exposed on the surface. Therefore, diamonds can not
make cracks. If you cut the wafer in this condition, big
chippings may probably be caused, or the blade may be
broken depending on the cutting speed.

After dressing, bonding material is removed and diamond
comes out on the surface as shown in the right sketch below.
At the same time, small hole which we call chip pocket is
created. This chip pocket will bring cooling water in the
cutting area and will draw out small cutting chips
temporarily storing in this pocket.

2.1.1 The Blade


Figure 6: Elements of blade structure and their purpose

The blade is composed primarily of grit and bond. The grit
is what actually performs the processing. The bonds role is
to the hold the grit in place.

2.1.3 Relationship of Blade Condition to Surface Chipping


Figure 7: Relationship of Blade to Chipping

2.1.4 Dressing and Pre-Cutting

Chippings are generally present on a new blade. So, blade
dressing and precut are need to be performed.

Figure 8: Dressing and Pre-Cutting Mode


Blades are dressed before shipment. However, precut
operation is still needed to condition the blade and to true
the outside diameter, removes excess binder material or
loose diamond particles, and minimize the load, creating a
cooler and freer cut resulting to minimize occurrence of
chippings.

2.1.5 Step Cutting Mode

Dressing and Pre-cutting cannot simply eradicate chippings
when using a single blade. Single blade carries a greater
process load and thus, results in an increase in surface
chippings. That is why a Step-Cutting mode was introduced
to minimize chippings during cutting. Step-Cutting method
is done using two blades (Z1 & Z2). The Z1 will partially
22nd ASEMEP TECHNICAL SYMPOSIUM
5
cut the wafer and Z2 will totally cut the wafer making it
stress relief.


Figure 9: Step-Cutting Method
2.2 Compression Mold (Critical Process #2)

One of the integral components in the production of IC is
the molding compound; a thermo set plastic epoxy use as
packaging material to protect it from external environment.
Unlike conventional transfer Molding, Scalable device
process uses Compression Molding with ultra fine filler
compound.


Figure 10: Compression Molding Mechanism

The advantages of compression molding system are
zero/less wire damage, good filling on narrow gap on die,
and no cull/no runner. The technology was necessary for
Scalable device due to the requirement of the mold cap of
0.28 mm or 280 microns thickness. With this, device is
prone to Voids during molding thats why Voids became the
top reject contributor.

Mold voids are commonly easy to correct, but this requires a
thorough parameter optimization through DOE. Design Of
Experiment was done to achieve desired parameter range for
molding process taking to account the critical input and
output responses. In this case, Mold Voids as the critical and
primary output response.

2.3 In-strip Test (Critical Process #3)

In traditional way, units are tested after singulation, but in
limited quantity. In this era of technological advancements
of high density device, In strip Testing was developed. The
dilemma however is the contacting issues. The scalable
device consists of 12,740 singulated units make it prone to
alignment failures compare to conventional device consists
of less than 500 units.


Figure 11: Singulated Units showing almost zero gap in
between dice resulting to AA (Auto Align) Failures


3.0 EXPERIMENTAL SECTION


3.1 Materials:
Assembly machine, based on identified three
critical processes;
o Wafer Sawing Machine
o Compression Mold Machine
o Probe/Test Machine
Scalable devices (Wafer/Lead frames)
Sawing Blades
Molding Compound
Probe Card

3.2 Design of Experiment

A DOE for Compression Mold was conducted with the
objective to determine and define window for critical
parameter range, thus eliminate Mold voids. Shown below is
the DOE matrix ran using SAS-JMP, system software
calculates automatically the combination of runs.


22nd ASEMEP TECHNICAL SYMPOSIUM
6

Table 3: 3x3 Full Factorial Design for Mold Voids

Full factorial design with a total of nine runs was created. At
SAS-JMP. Mold Temp and Cure Time were identified as the
most critical parameters that will cause mold voids defect.
Results of each run will be discussed in the results section.

3.3.The Reverse Flow

In order to eliminate alignment issues, an orthodox move
was made. The reverse flow which is testing prior
singulation will ultimately resolve Auto Align and other
singulation related defects as testing will be done on a strip
form.


Table 4: 4M+1E Matrix to identify and address AA Fails.
Highlighted factors are significant.

4.0 RESULTS AND DISCUSSION

Optimum process parameters were attained based on the
results of the DOE that will address the top reject
contributors to the critical processes.

4.1 Wafer Sawing Optimization to Address Chippings

Now that we see the relationship of blade to surface
chippings and the requirement to undergo dressing and pre-
cutting plus Step-Cutting mode to mitigate the chippings,
DOE on how these three input variables were performed.
See table below for the results.


Table 5: DOE Matrix using Dressing, Pre-Cutting and Step-
Cutting mode as input variables.

Comparative Tests were used to statistically validate the
results, with the aide of SAS-JMP, a statistical software
which greatly facilitates in analyzing the data and relieves
much of the tedious calculation. Tukey-Kramer test was
used for it gives a more conservative estimate of results as
compared to the other tests. Statistical graphs as shown
below in Graph 5, was also presented for ease of
interpretation and analysis.

Above DOE Results confirm that when blade is Dress, Pre-
cut and used Step-Cutting mode as shown in Table 4, gives
minimal surface chippings. The circle that was separated to
other four circles explains that RUN 5 denotes a significant
difference. Validations are made to verify effectiveness of
the Run 5. As shown in Tukey-Kramer test reveals a highly
significant difference in terms of surface chippings among
other runs.


Graph 5: Statistical Graph/Analysis of Variance showing
RUN 5 w/ Significant Difference in terms of Means and Std
Dev

4.2 Compression Mold Optimization to Address Voids

During development the initial problem encountered was
package voids every shot. Together with the Towa mold
machine supplier and the Mold compound supplier
(Supplier X), DOE was performed using a matrix of
different batch of mold compound and sets mold parameter.
The DOE result is illustrated in below figures.

22nd ASEMEP TECHNICAL SYMPOSIUM
7

Table 6: DOE Matrix to optimize Voids during
Compression Molding

DOE results of Compression Mold showed that optimum
parameters in terms of Voids can be achieved by using the
175 deg C and 180 sec Curing Time regardless of molding
compound used.


4.3 In-Strip Test Optimization to Address AA (Auto
Align) Fails
Table 7, below shows comparable yield and test results
during preliminary evaluations when reverse flow is
implemented without AA failures.


Table 7: Yield Comparison of un-singulated vs. singulated
units causing AA Fails.

Although preliminary evaluations were made, large scale
validation is needed as Reverse Flow is considered major
change and will undergo board review via Process Change
Review (PCR). It will take a longer time to implement due
to its major change requirements. AA fails was still further
investigated while waiting for the reverse flow to be put in
place. Cause and Effect matrix was tabulated to identify
other factors contributing to this defect.

After identifying the potential causes and validating its
contribution on AA Fails, the following solution and error
proofing was created. Solution was put in place based on
cost, applicability, effectiveness and impact to the problem.
During the course of brainstorming, a breakthrough idea
came out that will defeat all odds. AA fails will ultimately
resolve by reversing its process, as AA fails occur when the
products are singulated brought about by traditional way of
testing units after singulation, this time testing was done on
a strip form prior singulation thus eliminating the problem.


Table 8: Solution Validation Matrix

4.4 Verification of Results

After the implementation of the identified solutions, level of
rejections was monitored. Shown below in Graph 6 are the
results before and after the solution implementation.


Graph 6: Top Reject of Critical processes Trend before and
after the implementation of the Corrective Actions. (Actual
DPPM values intentionally not shown)

22nd ASEMEP TECHNICAL SYMPOSIUM
8
Significant reduction in the level of rejects were achieved
and the three critical processes and their corresponding top
reject contributor become stable after the implementation of
corrective action. A Poka-Yoke approach by reverse flow
lead to the elimination of AA Fails and a remarkable
improvement of 95% gained after the implementation of
corrective actions for Voids and Chippings through
comprehensive Design of Experiments.

This is a good indication of manufacturing preparedness for
full production mode.

4.5 Cost Savings
In a short period of time of implementation, Cost savings
was computed after confirming the effectiveness of the
implemented corrective action. A total of 9K USD is the
computed savings for the for the 3 quarters averaging 3K
USD per month based on the scraps/yield loss brought about
by unoptimized process conditions during ramp-up stage.
And this has a big impact in a cost savings drive of the
company and as benchmark for new product with less
production cost but with high quality standards.

5.0 CONCLUSION

Although, a flawless New Package Introduction cannot be
realized immediately, process optimizations play a vital role
to as early as line stressing stage, before full production
release can be granted. Employing in depth engineering
analysis and with the aide of statistical analysis in solving
top reject contributors were presented on this paper. Using
the knowledge and understanding on statistical tools lead us
to pinpoint the critical processes that need special attention
and focus during risk production. Top reject contributors
were identified using pareto analysis and problems were
addressed using Design-of-Experiment and solution
validation was employed to formulate effective corrective
actions. Chippings at Wafer Sawing can be addressed by
doing dressing, pre-cut and step-cutting mode. Voids
induced during Compression Molding can be eliminated
using optimum parameters via DOE of 175 deg C and 180
sec Curing Time and Auto Align failure can be eliminated
by thinking out-of the box idea by reversing its process.


6.0 RECOMMENDATIONS

It is recommended that the corrective actions identified, be
sustained and continue to monitor to maintain the rejects on
the acceptable PPM level as some of the identified rejects
cannot be zeroed out or eliminated. This project showed
how to dig and identify contributing factors on the top
rejects of critical processes by doing a during early stage of
production and employing in depth engineering and
statistical analysis to attain significant improvements and
recommends a permanent fix to production line. It is
imperative that when new devices are coming in, Critical
processes are needed to be identified and that appropriate
corrective actions and solutions be made so that when full
production are set, deliveries will not be at stake.


7.0 ACKNOWLEDGMENT

The authors would like to extend their sincerest appreciation
to the following: Willy Cabreros, Jun Bernabe, Addonyz
Antonio and the rest of Scalpack assembly.

To our beloved family, colleagues, friends, and co-
employees who gave their full support in making this
project successful.
Most especially to our God Almighty for giving us
continuous blessings in our career and everyday
undertakings.

8.0 REFERENCES

1. DISCO Technology handbook
2. DISCO Basic Dicing Blade Module
3. TOWA Automold Machine Manual
4. EG Pathfinder Prober Manual
5. ST Internal Documents

9.0 ABOUT THE AUTHORS





Antonio Dhon Sumagpang Jr. received his B.S. degree in
Electrical Engineering (BSEE) at Technological Institute of
the Philippines, Manila. He has 15 years solid experience in
Semiconductor Industry, 12 years of which acquired from
STMicroelectronics (Formerly NXP/Philips). He
participated in the 20
th
ANTS (ASEMEP National Technical
Symposium) last June of 2010, and won the Best Technical
Paper Award in Production Track Category. He has
numerous awards in Regional and National Quality
Competitions. Recently, he was transferred to New Product
Introduction as Sr. Process Engineer under the Assembly
Department of STMicroelectronics Calamba, a Quality
Practitioner and a certified Green Belter.





22nd ASEMEP TECHNICAL SYMPOSIUM
9
















Arnold Rada received his B.S degree in Mechanical
Engineering from Adamson University. In November 1999,
he joined Philips Semiconductor Philippines Inc. wherein he
spent hes 7-years of service as Mold Process Engineer for
Standard and Advance package product. On the year 2006,
he was transferred to NPI Engineering to handle
qualification of new materials as well as New Products
introduced to assembly manufacturing. On the year 2010, he
was involved in the start-up of the Scalpack package where
hes major activity focus on Initial production management,
that scope of which is to handle preparation of the line in
terms of manpower, material, documentation/methods,
process improvement and quality related issue.