Page 47 - ChipScale_Mar-Apr_2021-digital

is according to the industry standards The TC-J cycling condition
JEDEC TC-J condition [12], and IPC- applies a uniform temperature
9701A [13]. The thermal chamber used distribution across the entire
for the assessment is a Thermotron SE- package on board assembly,
600-10-10, single-zone chamber. Target and therefore only requires a
failure criteria is a 100 ohm shift from mechanical model to complete
time zero set points for greater than 50% the analysis. A total of six TC-J
of the hardware. thermal cycles (0°C to 100°C)
are run in the model to achieve
Modeling methodology stability in the plastic strain/
The simulation in this study aims to plastic work accumulated per
analyze and compare BGA reliability in cycle. For the field application
two different thermal cycling conditions. m o d e l , t h e t e m p e r a t u r e s
The first case is the JEDEC standard w it h i n t h e p a ck a ge a r e a
TC-J thermal cycle, which is uniform result of Joule heating inside
temperature cycling. Within this case, we the logic die and HBM. The
have two configurations: one with and one temperature distribution is
without a heat sink. The second case is a also nonuniform because of
representative field-like cycle. We call it a the different dimensions and
field-like cycle because the actual cycling thermal properties of the package
conditions usually depend on the customer components. Therefore, a steady
application. In this case, we analyze only state thermal model is solved
the configuration with a heat sink. This first to evaluate the temperature
condition is nonuniform temperature distribution across the entire
cycling and therefore it involves both assembly. These temperature Figure 3: TC-J cycle Weibull failure rate plot.
thermal and mechanical models. values are then input into the
3D FEM models are generated using mechanical model as thermal loads. The
ANSYS Mechanical software (Figure 2). temperatures evaluated in this model
The model consists of a fully-assembled correspond to the ON state of the module.
lidless 2.5D package with a stiffener ring, For the OFF state, a uniform temperature
printed circuit board (PCB), backing of 25°C c or r e s p ond i ng t o r o om
plate (when a heat sink is used) and a full temperature is assigned. The thermal
array of BGAs. The heat sink/pressure model is run only once for a given set of
plate component of the assembly itself is boundary conditions. Once the model is
not modeled and is instead replaced with completed, the same temperature values
an effective pressure or convection heat stored in the results file are read into
transfer coefficient. In order to model the structural model for each field cycle.
Three different Tj max (maximum junction Figure 4: Die and pry test showing failed BGAs
temperature) conditions of 100°C, at corners.
115°C, and 160°C are analyzed. These
correspond to different use conditions for assemblies with heat sinks is seen to
of the product in the field. A cycle time be nearly half the characteristic life for
of 24 hours with 11 hours dwell at ON, the assemblies with no heat sinks. Using
OFF states and 1 hour transition times is the 540-pin connector to isolate fails
considered. Similar to the TC-J cycling and further leveraging the dye and pry
model, a total of six field cycles are run technique to characterize failures, the
to achieve stability in the plastic strain/ presence of BGA fails was observed at the
plastic work accumulated per cycle. corners of the module (Figure 4). Using
Figure 2: Global FEM model of a 2.5D package on a this analysis, several corner failing BGAs
PCB with a backing plate. Results and discussion were detected with pad rupture observed
The following sections discuss results on surviving solder balls. Additionally,
the BGAs to a higher degree of accuracy obtained from TC-J reliability testing, 100% failure was observed for the heat
without being computationally expensive, TC-J cycling simulations, and field sink mounted hardware. There were no
a global-local modeling approach is cycling simulation. signs of electrical or physical failures in
utilized [9,10]. The material properties TC-J reliability test. The normalized the interposer or chip shadow regions of
in the global and local model are all Weibull failure rate plot for the two the laminate.
temperature-dependent elastic properties configurations is shown in Figure 3. A TC-J thermal cycling simulations.
except BGA solder. Anand’s viscoplastic clear difference is noted in the time to Equivalent plastic strain is used as a metric
relation is used to define the constitutive failure between the two configurations. to track the damage in solder. The location
behavior of solder. The 63.2% Weibull characteristic life (η) of the critical BGA is determined using

45
Chip Scale Review March • April • 2021 [ChipScaleReview.com] 45