Page 22 - ChipScale_Mar-Apr_2020-digital

to SACB. Usage of underfill was also
evaluated to improve FOWLP BLR-TC life.
The SACA/no UBM case leg had the first
fail at 369 cycles when no underfill is used.
However, when underfill is used, the first
failure was observed at 5,147 cycles, and
only two fails were observed by testing up
to 6,994 cycles. When a stiffer solder alloy
is used, the BLR-TC failure mode often
moves from the solder to the package RDL
and silicon die BEOL cracking [6,7]. To
overcome these package reliability risks
across different levels of interconnects,
the RDL stacks’ designs were optimized
Figure 6: Weibull plots from BLR-TC test for different DOE legs for FOWLP: SACA no UBM, SACA UBM, SACB no for the different solder alloys and no UBM
UBM, and SACB UBM. and UBM metallizations. This resulted
in improved package reliability for all
coefficient of thermal expansion (CTE) Figure 6 shows the experimental BLR-
mismatch between package and PCB, which TC data for the different DOE legs: SACA the different interconnect levels ranging
in turn stresses the solder ball interconnect no UBM, SACA UBM, SACB no UBM, from the silicon BEOL, the package RDL
between the package and PCB. Typical and SACB UBM. The Weibull plot shows interconnect, to the package/PCB solder
failure modes are depicted in Figure 1. For SACB solder alloys have significantly joint. In addition to BLR-TC, AL-TC was
fan-in wafer-level packages, this stress is improved FOWLP BLR-TC performance. also performed where the device product
most severe where the Si is bonded to the Table 4 lists the characteristic life cycles for was mounted on board and underwent
substrate because there is a large difference the different DOE legs. SACA and SACB temperature cycling. Full product functional
in CTE between the Si and the PCB (~3ppm alloys with UBM have between 23%- 29% testing (FT) was used to detect failures. No
for Si and ~15-20ppm for the board). The improvement in characteristic life versus FT fail or product performance degradation
FOWLP body is a composite of silicon no UBM. For no UBM, changing the solder was observed up to 1500 AL-TC cycles.
plus the mold compound, so the package alloy from SACA to SACB leads to a 341% Figure 7 shows drop test Weibull plots
CTE is relatively higher than that of only improvement in BLR-TC characteristic for SACA no UBM, SACB no UBM, and
silicon. Still, however, the CTE mismatch is life. For the UBM case, the improvement in SACB UBM. SACB solder had significantly
quite large between package and PCB [10]. BLR-TC characteristic life is 363% when better drop test reliability performance
During temperature cycling, creep of the the solder alloy is changed from SACA compared to the SACA alloy. For the SACA
solder interconnects occurs at the elevated no UBM leg, the first fail is observed at 35
temperature of the thermal cycling. drops. A sample failure rate of 50% was
The mean characteristic life during BLR- observed when testing up to 500 drops.
TC can be modeled and predicted by either For SACB solder alloys, drop testing was
the accumulated creep strain energy density, done at a level of up to 1000 drops, and
or the accumulated creep strain model only one fail was observed for no UBM
[11,12]. Accumulated creep strain and creep Table 4: BLR-TC characteristic life (~63% fail) for and UBM legs at 432 drops, and 679 drops,
strain are direct representations of the creep different solder alloys and no UBM, UBM case. Values respectively. Figure 8 shows the bump
damage that the solder joint undergoes in parentheses indicate relative characteristic life shear strength comparison between SACB
during BLR-TC. For the same package values, with SACA no UBM as the reference. and SACA solder alloys at time 0, and after
structure, using a stiffer and stronger
solder alloy and UBM results in significant
improvement in mean characteristic life
[5]. In this work, four DOE legs were
built with SACA and SACB solder alloys
with, and without, a UBM. Table 3 shows
the properties of solder alloys that were
evaluated. Greater tensile strength of the
solder has been known to increase the
lifetime of the solder during cyclic bending
in BLR-TC testing [6].

Table 3: List of material properties of investigated
solder alloys. Figure 7: Drop test Weibull plot: a) ∆ – SACA/no UBM; b) - SACB/no UBM; and c) € - SACB/UBM.
∆
20
20 Chip Scale Review March • April • 2020 [ChipScaleReview.com]