Chip Scale Review September • October • 2018[ChipScaleReview.com]
The clouded view of WCSP inspection strategies
By M. Todd Wyant
emiconductor parts continue
to shrink at alarming rates.
As par ts approach g reater
than one million units on a three hundred
millimeter (mm) analog wafer, many visual
containment actions are being pushed past
the breaking point. System improvements
for detection are limited as algorithms, and
detection strategies are still limited when it
comes to small defects on traditional flows.
shows the progression from 1mm
part sizes down to 0.125mm, highlighting
the difficulty encountered when trying to
detect defects on such small components.
Today, more than ever, there is an
increased pressure to ship parts without
defects and achieve single-digit defective
parts per million (dppm) levels. The
continued industry push is to apply new
technology and methods to ultimately
create a situation where all defects are
contained or eliminated prior to shipment.
Today’s inspection equipment can scan
par ts quick ly and produce det ailed
inspection reports to drive improvements,
but with part sizes exponentially shrinking
and key return defect modes not shifting,
the push now is to f ur ther improve
inspection capabilities to detect problems
even with the smallest of components.
The cont i nued pu sh t o i nc r e a s e
resolution has reached the tipping point.
Resolution is becoming so detailed that all
components are starting to look different.
The variation being introduced with this
high resolution is leading to false rejectivity
that is hindering the situation and time
needed to drive improvements (
Implementing filtering and other methods
to prevent false rejectivity is leading back
to release of the same returned items into
the pack solutions that ultimately end up
back into the customer’s hands. In addition,
this complexity drives engineering time
and effort that ultimately impacts cost.
This balance between over-rejection
and rejected shipments has been a long-
standing problem, and demonstrates the
real challenge of implementing a quality
assu rance system. Meeting moder n
manufacturing demands of delivering
fast and accurately is challenging, while
simultaneously driving cost through scrap
reductions and improved process efficiency.
Current working path
Today, assembly sites are driving to
implement inspection solutions that use
infrared (IR) or standard high-resolution
imaging to detect, measure, and contain
defects. These methods rely heavily on
lighting, imaging and software to detect and
contain rejectivity. However, the issue with
this method has shown that false rejectivity
will be a way of life. During testing on
many different platforms, anywhere from
2-8% of false rejectivity has occurred
with the best case settings. Depending
on the system type, productivity loss can
be anywhere from 5-50% for increased
inspection time required.
T h e n e w m e t h o d s h a v e a l s o
demonstrated limitations in finding and
detecting all rejectivity. In addition, as
, the use of rotated silicon
structures prevent IR from passing through
the substrate during visual inspection. In
the system during inspection, the silicon
lattice prevents clean exit of the infrared
light penetration through the component.
The result of IR on these products is a black
surface that is not visible with detection
software. The chart shows the difference
with varying substrate rotation, with and
without backside coatings below, and
highlights the challenges of inspection
using IR technology solutions.
Challenging defects are really what is
driving inspection capability increases,
while cost and end users continue to drive
shows the dilemma
with any inspection process as defect sizes
are very small. Cracks and damage cross
This graphic shows the progression from
1mm-part sizes down to 0.125mm, highlighting the
difficulty encountered when trying to detect defects
on such small components.
Examples of challenging rejects: a) Side
wall crack; b) Corner crack; c) Common edge chip; and
d) Hairline cracks.
Rotated silicon effects on IR transmissivity.
Examples of marginal defects: a) Whisker crack; b) Corner hairline crack; and c) Crack failure limit.