Page 50 - ChipScale_Mar-Apr_2020-digital
P. 50
Process characterization: design
of experiments
Since early 2016, numerous studies
have been conducted on actual packages
to makes sure that LAB data does
indeed indicate a substantial benefit
over standard MR and TCB processes
before ramping into mass production.
In terms of the assembly process, there
is no difference between MR and LAB
except that the MR reflow process is
interchanged with LAB (Figure 7). Both
capillary underfill (CUF) and molded
underfill (MUF) can be used in the
Figure 5: IR laser time before and after solder joining. LAB process. LAB process substrates
don’t need to be heated to a very high
temperature like what is needed for MR,
therefore, it only takes a few seconds to
heat up the die and perform chip joining.
In one of the early studies, two
TVs (one flip-chip fine-ball grid array
[fcFBGA] and one flip-chip ball grid
array [fcBGA]) were selected for LAB
process qualification. Fine-pitch Cu pillar
bump with bond on lead (BOL) pads were
introduced in the substrate design. One
of the great advantages of a BOL pad
is that it allows for very relaxed design
Figure 6: Bump joint displacement data between MR and LAB. Data are measured at die left, center, and right locations.
Figure 7: Typical assembly process difference
Table 2: Bump offset /displacement data comparison between MR and LAB processes. between MR and LAB for a fcFBGA molded package.
predictions are listed in Table 2. Empirical
displacement data was also collected from
the actual package cross section after the
chip attach process. A very small bump
offset or displacement was observed in
LAB over MR as shown in Figure 6.
Absolute displacement numbers between
calculations and empirical results may not
match, but trend wise, good correlation has
been established between calculations and
empirical data as illustrated in Table 2 and
Figures 5 and 6.
Table 3: TV package attributes for the LAB study.
48
48 Chip Scale Review March • April • 2020 [ChipScaleReview.com]
