Page 10 - ChipScale_Sep-Oct_2020-digital

EXECUTIVE VIEWPOINT

Automotive – Driving Zero Defects

Chip Scale Review asked David F. Hanny, Director of Marketing at Applied Materials, Automation
Products Group, to provide insight into how market growth in advanced driver assistance systems
(ADAS), electric vehicles (EV), and autonomous vehicle (AV) technologies is driving the need for a
zero defects strategy in the manufacture of integrated circuits (ICs).

CSR: Because ADAS/EV/AV market
growth is raising the complexity of ICs—
as well as how they are used in systems that
must make almost instantaneous decisions
in traffic situations—what are the most
significant limiting factors with respect to
achieving a zero defects strategy in their
manufacture? How can you overcome those
limiting factors?
DH: We see three primary limiting
factors on quality as we move towards
zero defects in manufacturing. First is the
slow development of new technologies and
materials for new product introduction.
In automotive chip manufacturing, new
product yield begins as low as 40% for
a period before it moves up to typical
yields in the 88-92% range. Next is the Figure 1: An illustration of end-to-end quality. Moving decisions from offline human decisions to real-time
introduction of new raw materials, along decisions based on data patterns enables factories to increase product quality.
with a third factor being errors in human
decisions. Each are inhibitors of quality in technology (SMT) lines are beginning AI techniques are being developed and
the fab. These challenges can be addressed to increase their automation capabilities. improved upon from earlier approaches?
with increased requirements, measures, At least one major original equipment DH: Factories operate in varying degrees
and validation over supplier materials manufacturer (OEM) has increased the of automation from operator-driven to the
and quicker learning cycles of anomalies. automation requirements on their packaging early stages of full automation (see phase 3 in
Moving decisions from offline human suppliers, requiring them to add more sensor Figure 2). Many companies are challenged
decisions to real-time decisions based on monitoring (like fault detection) to maintain to have the kind of end-to-end quality to
data patterns enables the factory to increase their status as a valued supplier. A common leverage AI. The primary reasons are due to
product quality (Figure 1). trend for packaging and SMT lines is seeking the economics and infrastructure of today’s
for more advanced quality capabilities. factories. Chips that are highly specialized
CSR: How can the industry improve for automotive applications such as ADAS
the way field failure data is married to CSR: What role is artificial intelligence and light detection and ranging (LIDAR) are
quality issues with respect to semiconductor (AI) playing in the end-to-end quality typically manufactured in 300mm fabs that
processes in the fab or at the outsourced chain? Can you describe in more detail how have infrastructure and systems running at
semiconductor assembly and test (OSAT)/
packaging supplier?
DH: The real challenge with performing
failure analysis is that it relies heavily on
the genealogical granularity of the data
throughout the supply chain. Not every
factory has the same level of ability to
diagnose, and the process can be very
manual. 300mm factories have developed
more tools and have greater access to this
data. Often the node in the supply chain
that can’t afford to answer the question gets
stuck with the bill. To combat this challenge
many packaging factories and surface mount Figure 2: The roadmap to full automation: Factories operate in varying degrees of automation from operator-
driven to the early stages of full automation.

8 8 Chip Scale Review September • October • 2020 [ChipScaleReview.com]