Achieving Fine-pitch Ball Placement

[SMT & PACKAGING]
Achieving Fine-pitch Ball Placement
Demands to reduce the dimensions of even the most highly miniaturized
semiconductor packages are placing pressure on IC vendors and
packaging specialists to develop faster and more productive packageassembly processes. Ball placement techniques are being fine-tuned to
address these issues.
TOM FALCON, DEK
S
ince the wafer-level chip-scale package
(WLCSP) has become the package
technology of choice for devices such as
DRAMs, flash memories and small
FPGAs, research has focused here to find
suitable miniaturization measures. The next
evolutionary step for WLCSP is to reduce the ballgrid interconnect array pitch to 0.3 mm, which
allows semiconductor manufacturers to increase
the number of die per wafer by more than 50%, for
little extra manufacturing cost. However, continued
shrinkage of pitches and ball diameters brings
challenges of implementing reliable and
repeatable processes for placing 0.2-mm solder
balls on a 0.3-mm pitch at high yield and
throughput.
screen-printed onto all solder-bump pads. After flux
printing, the wafer and pallet are transported to the
ball-placement machine. Following accurate visual
alignment with the metal ball-placement stencil,
the wafer is brought into contact with the underside
of the stencil before the ball transfer head traverses
the topside of the stencil. This deposits a single
solder ball into each of the stencil's apertures
(Figure 1). The alignment process ensures that the
apertures coincide accurately with the fluxed
solder-bump pads.
Direction
of travel
Ball place
transfer head
Ball Placement Techniques
Over the last 10 years, several methods for the
mass transfer of solder balls to wafers has been
developed, as well as complementary processes for
rework and single-ball placement. An accurate and
repeatable stencil printing ball placement process
for 0.2-mm solder balls on a 0.3-mm pitch was
developed, with the goal of demonstrating greater
than 99.99% placement yield at a throughput of
60UPH or more. These are the fundamental criteria
for a ball-attach process supporting cost-effective
mass production of 0.3-mm pitch WLCSPs.
Equipment and Process Overview
Solder-ball placement using screen printing begins
with tooling the wafer in a conveyorized aluminum
pallet. This pallet is transported into the fluxprinting machine where the wafer is visually aligned
with an emulsion-mesh fluxing screen with
reference to laser-marked fiducials. Flux is then
Pallet
Vacuum
channels
Wafer
Ball place
stencil
Fig. 1. Solder ball placement.
Screen and Stencil
The specification of the emulsion-mesh screen used
for the fluxing process is derived from established
formulas based on the pitches and ball diameters to
be placed. Typically this will be a stainless-steel, 45°
mesh with several microns of photoimageable
emulsion on the wafer side. The screen's mesh wire
diameter, mesh pitch, and emulsion aperture size
are calculated according to the ball diameter to
achieve the optimal ratio of printed-flux volume to
solder-ball volume. Once fabricated, the screen is
checked against specifications and parameters
including aperture size, mesh tension, emulsion
[SMT & PACKAGING]
integrity, and image-stretch or shrinkage.
The ball placement stencil is composed of two
layers. The top layer is either stainless steel or
electro-formed nickel, and the bottom (or standoff) layer is a photoimageable dry-film resist. The
purpose of this stand-off layer is to hold the top
metal layer away from the pre-printed flux. Another
benefit of this material is that it provides a relatively
soft surface in contact with the active side of the
wafer, thereby minimizing any possibility of wafer
damage.
Wafer Tooling
The wafer tooling is a two-part system consisting of
a universal base plate, with a user-configurable
series of vacuum rings suitable for wafers from 100
to 300-mm.
This is coupled with a top plate, or shim, unique to
the wafer diameter and thickness. Generally, the
shim aperture in which the wafer sits is 1-mm
greater than the wafer diameter, and the shim
thickness is lower than the wafer thickness, but by
no more than 50 µm. This creates a smooth and flat
platform to support the metal stencil and solderball transfer head during the ball-placement
process.
Experiment
Having prepared all equipment to a greater level of
accuracy than is normally required, the placement
experiment commenced. To verify achievement of
the desired throughput of 60 UPH and ball yield of
99.99%, an entire cassette of 25 wafers was
processed, with no operator intervention and at a
cycle time of 60 seconds or less.
Flux Print
Flux printing for ball placement uses established
screen printing techniques coupled with state-ofthe-art printing machines and precision mesh
screens to create a robust process with a wide
operating window. Provided the mesh screen is
made correctly and the flux is well formulated for
screen printing, it is usually not difficult to find a set
of process parameters that produce excellent
results. In most cases a print gap of 1.5 to 2 mm,
and a print speed of 25 to 50 mm/sec, with a print
pressure of 2 kgs per 100 mm of (60 degree
polyurethane) squeegee blade length will produce
uniform flux deposits (Figure 2).
Equipment Modifications
The placement process requires that the wafer, its
pallet, the stencil, and the transfer head are all flat
to within a fraction of the ball diameter. Clearly any
reduction in ball diameter makes this even more
difficult to achieve. For this experiment great care
was taken to ensure these parts surpassed their
specifications.
Wafer tooling was checked with a dial-gauge
indicator and was within 30 µm. This was well
inside the normal specification of 50 µm in the
wafer area. The shim was matched to the wafer
thickness within 20 µm rather than the usual 50 µm,
and the base of the ball place transfer head was
precision-ground to within ±10 µm, in contrast to
the usual ±25 µm.
Taking these extra precautions ensures that the
wafer, stencil, and transfer head are all co-planar
within a few microns of each other over their entire
area, thereby enabling the smooth and accurate
transfer of solder balls from the transfer head,
through the stencil, to the wafer. Failure to ensure
this level of flatness in each element of this stack
could promote placement yield problems later on;
particularly damaged or sheared spheres.
Fig. 2. Typical flux deposits on blank 200-mm wafer.
This process is generally reliable enough not to
warrant a separate inspection stage. Established
practice is to perform inspection after ball
placement. The same protocol was followed for this
experiment: following initial setup, no flux-print
inspection took place. Interestingly, none of the
defects later captured by the ball place inspection
could be directly attributed to a flux printing defect.
Ball Placement
It is feasible to take a batch-processing approach
using this equipment, and flux all 25 wafers before
subsequently performing ball placement on each
[SMT & PACKAGING]
fluxed unit in the complete 25 wafer cassette.
Alternatively, units in one cassette could be fluxed
simultaneously with ball placement on a previously
fluxed cassette. However, since the equipment used
in this experiment was not configured for high
volume manufacture, all three operations (flux
print, ball placement, and inspection) were
performed sequentially for each wafer, before the
next wafer was processed.
In the industry generally, ball placement inspection
is sometimes performed post-reflow. However, in
this case it was carried out post placement so that
no reflow-generated defects could influence the
data. Therefore the placement process alone was
measured. Inspection was performed manually
and with both stereo zoom and video measuring
microscopes.
In this test configuration,
the machines are not
capable of 60 UPH,
because they are
configured for manual
loading. Automated
Fig. 5. 200mm wafer populawafer handling options ted with 269,108 0.2mm
are available that have solder balls on 0.3mm pitch.
already been proven to
meet 60 UPH for 0.3-mm ball diameters at this
wafer diameter. It follows that 60 UPH would also
be possible for this 0.2-mm process. Figure 4
shows the wafer cassette and ball-placed wafers
part of the way through the run. Figure 5 shows a
complete wafer immediately after ball placement.
Results
Conclusion
The results of the wafer-processing experiment are
shown in an SPC control chart (Figure 3). Defects
recorded were missing, misplaced, or damaged
balls.
This experiment has proven that it is possible to
successfully automate the placement of 0.2-mm
balls for WLCSPs with ball yield greater than
99.99%. It has also proven that the process is
relatively stable and controllable, with no data
points below the lower control limit, and no more
than four continuous data points either above or
below the mean yield.
By recording the machine cycle time and
comparing it with current high-volume equipment
installations for larger balls, we can also conclude
that the throughput requirement of 60 UPH can be
met for this ball diameter, provided robotic wafer
handling is used.
References:
1. M. Whitmore, M. Staddon, D. Manessis:
“Development of a Low Cost Wafer-Level Bumping
Technique.” International Wafer-Level Packaging
Conference 2004.
2. M. Whitmore, M. Staddon, D. Manessis: “The
Development Of Balling Technologies For Wafer
Level Devices With Pitches Down To 0.4mm.”
International Wafer-Level Packaging Conference
2005.
% Ball yield
Ball yield
Arithmetic mean
Upper control limit
Lower control limit
100.002
100.000
99.999
99.996
99.994
99.992
99.990
99.988
1
5
10
15
Wafer number
Fig. 3. SPC analysis.
20
25
Author Profile
Fig. 4. Ball-placed wafers in cassette.
Tom Falcon, senior process development specialist,
may be contacted at DEK Printing Machines Ltd,
Granby Idustrial Estates, Weymouth, Dorset, DT4
9TH, U.K.;+44/1305 208415; [email protected].