Page 15 - ChipScale_Sep-Oct_2020-digital

Moving these traditional good practices perspective. Process
to a holistic approach requires some p r obl e m s s u ch a s
integrated infrastructure. The first step is to whiskers and bridging
define what holistic means in this context. will elude many of
Holistic is the ability to understand the these tests, causing
interdependent behavior of the process shorts in the field.
steps. The control plan illustrates the So, the automation
known measurements associated with system provides users
any product. The automation system will the ability to define
need to provide users with the ability qualitative stack-up of
to define interdependency relations of step attributes, which is
different process steps, as outlined in the a measure of acceptable
control plan and process failure modes v a r i at i o n f or a
and effects analysis (FMEA). At runtime, given problem.
the system will consume the data coming It is true that other
from multiple process steps, as defined factors play a role in
in the control plan, and outline which the overall failure rate
data does not fit the population that we including electrostatic
expect for the specific processes. This damage and other forms Figure 3: Moving to holistic and intelligent systems.
will need to be done using real-time data of mishandling that can
tools because a single decision will require occur in several places
users to assess 15 parameters with 12 to throughout the supply
20 sites for each. Each site will need to chain. This highlights
have a statistic and be broken down to the need to have rapid
divulge within-wafer variance profiles, genealogy of quality
wafer-to-wafer variation within the same attributes that include
process step, and finally, variation inherited design. Design is the
from upstream steps. This approach will concept of parts that
reduce the variation arriving at final have been made for a
test and therefore, reduce the chance of long time but are now
failing parts leaving the facility. More used in a new way
important is the change that this approach that is not suitable
will impose on final test validity. A more for their reliability.
stringent variation verification will become This constitutes an Figure 4: Striving for zero defects.
more effective to capture a greater degree approximately 0.15PPM
of nonconformance. contribution to failures for any given device and cause. The speed
Another key difference is that holistic in the field. The cases that are “unknown” of resolution in this case will impact cost of
systems are not only error driven, but represent an approximately 0.21PPM liability and more importantly, safety. Unless
also sensitive to variation within the contribution to the failures and are a major the design of new automation systems adopts
specification limits. Data show that this liability to the vendor. In the case where no these guiding principles, it will be difficult
is happening in facilities that have yields cause can be assigned, the small vendors will to expect any supply chain to converge
ranging from 88% to 92%. The front- be held accountable for the failure and will effectively towards zero defects (Figure 4).
end fabs combined with the packaging have the cost deducted from their agreement
operations account for an approximately with the big car manufacturers. If this ability Reference
0.76PPM failure rate. Devices within could be automated to straddle the front Original Equipment Suppliers Association
the specification limits will ultimately end and back end and ultimately include the Page 54 - https://www.oesa.org/sites/default/
be functional, but not necessarily the surface mount technology (SMT) line, then files/automotive-defect-recall-report-2019.pdf
same from a performance or longevity rapid identification would become possible

Biographies
Selim Nahas is the head of the Process Quality Group, which is part of the Applied Materials, Automation
Products Group. He has 25 years of experience in Semiconductor factory automation systems including
Material Execution Systems, Automated Material Handling Systems and Advanced Process Control.
Selim holds degrees in Industrial Engineering and Arts and Sciences from Northeastern U., Boston, MA.
Email Selim_Nahas@amat.com
Mana Dedhia is a Quality Engineer at Analog Devices, Inc., Wilmington, MA. He was born in India and
completed his Bachelors in Electronics Engineering from the U. of Pune and his MS in Electrical Engineering
from the U. of Minnesota, Twin Cities. He started his semiconductor journey at Globalfoundries in Fab 8, Malta, NY in 2012.

13
Chip Scale Review September • October • 2020 [ChipScaleReview.com] 13