16 Processing and Reliability of Solder Interconnections …
491
approach towards minimizing an over stress or excessive TMF of the solder joints,
is the development of a controlled application process that dispenses the slush layer
in a repeatable manner.
16.2.6 Warpage Effects
Probably the most significant challenge that has faced the execution of stacked
package technology is warpage. Warpage can pose a challenge to both the assembly
process as well as the assurance of solder joint reliability in service conditions that
include temperature cycling. The latter scenario is discussed in the following section.
Between assembly yield and reliability, warpage has its greatest impact on the
former. This premise stems from the fact that the solder joints, being in the molten
condition, have zero load-bearing capacity. As a result, the shapes that are taken up at
the time solidification are controlled by (1) the molten solder surface tension and (2)
the warped geometries taken by the individual packages and PCB. Package warpage
determines the yield of the first-level interconnections, while warpage of the bottom
package and the underlying PCB control the second-level solder joint yield.
The schematic diagrams in Fig. 16.10 illustrate the two types of warpage experienced by packages and the potential defects. The top illustration shows the baseline
condition. Below it is the “frowny face” warpage, which can push together the outer
rows of molten solder so as to create a solder bridge (short circuit). The bottom
image is that of the “smiley face” warpage, which pulls apart the outer solder joints,
potentially leading to a gap (open circuit).
There are two caveats that accompany these diagrams: First of all, only the top
package has been allowed to warp. Solder joint defect formation becomes more
complicated when all of the packages, and the PCB, show a measure of warpage.
Secondly, the illustrated warpage occurred in only two-dimensions. There is also
the third dimension into/out-of the image plane, which can further complicate any
ability to predict defects in the solder joints.
Numerous studies—if fact, too many to describe each in detail here—have examined various factors that affect warpage. For example, Smith, et al. recognized that
the top package in a PoP stack has the greater priority in terms of managing warpage
because of the limited quantity of solder in the first-level interconnections [18].
Moreover, the same authors recognized that warpage is dynamic; it changes with
temperature as the stacked package progresses through the reflow profile. Relative
displacements can reach upwards of 100–150 μm at the solder balls, which is sufficient to cause the defects illustrated in Fig. 16.10. Zhao and co-workers explored
the effects of two factors, substrate design and top package design, on the latter’s
warpage behavior [3]. Copper density and balance in the substrate, as well as the
relative volumes and footprints of the dice and molding compound, are critical factors
with respect to package warpage. Similar trends were corroborated by Vianco, and
co-workers, using a computational modeling routine [8].
491
approach towards minimizing an over stress or excessive TMF of the solder joints,
is the development of a controlled application process that dispenses the slush layer
in a repeatable manner.
16.2.6 Warpage Effects
Probably the most significant challenge that has faced the execution of stacked
package technology is warpage. Warpage can pose a challenge to both the assembly
process as well as the assurance of solder joint reliability in service conditions that
include temperature cycling. The latter scenario is discussed in the following section.
Between assembly yield and reliability, warpage has its greatest impact on the
former. This premise stems from the fact that the solder joints, being in the molten
condition, have zero load-bearing capacity. As a result, the shapes that are taken up at
the time solidification are controlled by (1) the molten solder surface tension and (2)
the warped geometries taken by the individual packages and PCB. Package warpage
determines the yield of the first-level interconnections, while warpage of the bottom
package and the underlying PCB control the second-level solder joint yield.
The schematic diagrams in Fig. 16.10 illustrate the two types of warpage experienced by packages and the potential defects. The top illustration shows the baseline
condition. Below it is the “frowny face” warpage, which can push together the outer
rows of molten solder so as to create a solder bridge (short circuit). The bottom
image is that of the “smiley face” warpage, which pulls apart the outer solder joints,
potentially leading to a gap (open circuit).
There are two caveats that accompany these diagrams: First of all, only the top
package has been allowed to warp. Solder joint defect formation becomes more
complicated when all of the packages, and the PCB, show a measure of warpage.
Secondly, the illustrated warpage occurred in only two-dimensions. There is also
the third dimension into/out-of the image plane, which can further complicate any
ability to predict defects in the solder joints.
Numerous studies—if fact, too many to describe each in detail here—have examined various factors that affect warpage. For example, Smith, et al. recognized that
the top package in a PoP stack has the greater priority in terms of managing warpage
because of the limited quantity of solder in the first-level interconnections [18].
Moreover, the same authors recognized that warpage is dynamic; it changes with
temperature as the stacked package progresses through the reflow profile. Relative
displacements can reach upwards of 100–150 μm at the solder balls, which is sufficient to cause the defects illustrated in Fig. 16.10. Zhao and co-workers explored
the effects of two factors, substrate design and top package design, on the latter’s
warpage behavior [3]. Copper density and balance in the substrate, as well as the
relative volumes and footprints of the dice and molding compound, are critical factors
with respect to package warpage. Similar trends were corroborated by Vianco, and
co-workers, using a computational modeling routine [8].