Page 53 - ChipScale_Jan-Feb_2021-digital
P. 53
structures (Figure 1c), by applying a full
dose at the via and a partial dose for the
redistribution layer (RDL) lines, a single-
exposure process is used without impacting
throughput. The inherent flexibility of
maskless exposure, in conjunction with
the light valve’s ability to modulate the
exposure dose can reduce the number of
process steps required and increase yield
while maintaining throughput.
Dynamic data generation
Direct imaging systems have a wide
range of functions to increase flexibility for
advanced packaging and MEMS fabrication
processes, many of which utilize dynamic data
generation. Consider three different alignment
methods. The first and simplest of these is Figure 2: High-throughput compensated exposure process flow.
global alignment, which measures two or four
alignment marks on the substrate and then the substrate. Suitable for many processes – die-first processes, reconstituted substrates
adjusts offsets in the X, Y, and θ directions. including panel circuit board (PCB) – local exhibit random die offsets that require
Global alignment uses dynamic data alignment measures up to 400 points on a compensated exposure to achieve acceptable
generation to compensate for linear expansion/ substrate, calculates the difference between overlay performance. Steppers use a chip-by-
contraction and is particularly well suited for design position and actual position, and chip approach, measuring and exposing each
rigid processes, such as silicon or glass-based dynamically adjusts the exposure data for the die individually, but suffer from significant
substrates. Next is local alignment, which entire substrate. The third method – die-by- decreases in throughput as a result. On the
expands on the global alignment capabilities die alignment – is crucial for die-first panel/ other hand, direct imaging tools can expose
by compensating for nonlinear distortion of wafer-level packaging processes. In these all the die with one scanned exposure, using
51
Chip Scale Review January • February • 2021 [ChipScaleReview.com] 51