Miniaturization has driven the design of logic and memory integration using vertically integrated devices in a stacked configuration, where the top and bottom BGAs are mated to form Package on Package (POP) assemblies (Figure 1-1). The advantages of such a design, beyond the reduction of the packaging footprint, is to incorporate functionality and cost benefits for electronic products, such as cell phones, PDAs digital cameras, and other handheld devices. The intrepid nature of this technology has spread beyond the package realm, spawning new stacked die and SOC (System on Chip) technologies that integrate a similar vertical stacking approach, while retaining the needed functionality associated with more conventional designs.

The advantages of the POP architecture are such that one step reflow can be utilized in the concurrent processing of both the top and bottom BGAs. This approach to assembling POP BGAs utilizes a conventional paste process for the bottom BGA and a flux transfer dip for the top BGA. The top side of the bottom BGA, where the pads are typically a plated NiAu finish, provides a wettable surface pad area for the top BGA. The reflow process employed will depend largely on the composition of the solder paste, and the top and bottom BGA alloy. In situations where the paste alloy and BGA balls are of varying compositions, the reflow parameters should be adjusted to achieve the liquidus temperature of the higher melt alloy.

For the purpose of exploring various process options, the EMPF is conducting an experiment which will eventually produce thermal cycling reliability data on POP assemblies. Utilizing a mixed and all lead-free SAC 305 solder composition (Sn96.5Ag3.0Cu0.5), the experimental matrix will also be designed to include assemblies that will be underfilled, along with the more conventional approach of non-underfilled POP assemblies. The underfilled POPs will be further divided into bottom package only underfill, and top and bottom underfill. Typically, most POP architectures that integrate underfill as part of their process, underfill only the bottom package.

One of the potential outcomes of the experiment the EMPF was commissioned to produce was an attempt at differentiation in reliability under thermal cycling conditions. The experiment was designed to use various materials and processes that would be encountered in the manufacturing of POPs. The materials of choice were selected for their ability to withstand Pb free reflow temperatures. The solder pastes were composed of no clean flux and solder balls with a particle mesh of 300. The material used for underfilling was a high temperature, low CTE polymer, usually consigned for flip chip applications and ESS (Environmental Stress Screening) conditions where CTE is critical. The set of materials is listed below.

Solder Paste and Underfills used for the experiment:

SnPb                 Multicore MP200  No clean Solder Paste
SAC 305           Multicore LF320 – No clean
Underfill          Hysol FP 4548FC

The test vehicle for POP processing included a daisy chained PCB with an OSP (organic solderability preservative) coating, and top and bottom BGAs composed of SAC 305 alloys. The Bottom BGA utilized a top pad finish of Au/Ni. The vehicle contained 15 independent daisy chain sites where the top and bottom BGAs continuity could be tested for open circuits while isolating the particular location and BGA on the PCB. 

• PCB200 -OSP     
  A Daisy chained printed circuit board with an OSP coating.
  
  
• Top Package –A-FBGA
  152.  A 152 ball BGA with 0.65 mm pitch on a 14 mm square package footprint. The balls are comprised of SAC 305 solder.
  
  
• Bottom Package – A-PSvfBGA353 
  a 353 ball count with 0.5 mm pitch on a 14 mm square package footprint. The balls are comprised of SAC 305 solder.

Fixtures
The screen stencil for the solder paste was designed to apply a suitable amount of paste to obtain adequate wetting and contact between the BGA and substrate.  A flux transfer plate was utilized for assembling the top BGA to the top pads of the bottom BGA. The tacky flux was appropriate for Lead-Free processing and used the flux transfer plate as a shallow reservoir designed to hold a volume of flux that comprised ½ the diameter of the top BGA balls. The dimensions are below:

• Top BGA- Screen printing Stencil at a thickness of about 0.12 mm with an opening of 0.28 mm.
  
  
• Bottom BGA Application - Flux transfer Plate machined to a depth of ½ ball diameter, approximately 0.150 mm height.

Experimental Matrix
Table 1-2 illustrates how the 112 assemblies processed for the experiment were segregated and what manufacturing process and materials were chosen for each.

Manufacturing process
The assembly process identified in the manufacturing process matrix (Table 1-3) with an “X” indicates the operation has been performed for the specific manufacturing process. As an example, the yellow shading highlights the manufacturing differences, specifically where the underfilling process was concerned. The underfilling for manufacturing process “3-6”,   incorporated a two step reflow that allowed the bottom package underfill to reflow after the top package assembly. The manufacturing process “7-8”, underfilled the bottom package after the assembly of both top and bottom packages.

Some initial results:

Processing Issues
Data collected thus far indicates that any reoccurring failures are attributed to missing solder balls. When this issue was discovered, particular attention was given to the inspection of the BGA’s for missing solder balls, so the packages that had dislodged balls were separated. Even after the segregation process, and subsequent step by step inspection, opens were occurring occasionally on different sites due to missing balls. A particular concern during the processing of POP assemblies, is the propensity of package warpage that would result in open solder joints, particularly in the bottom BGA. Co-planarity of the bottom package against the substrate becomes an issue as the corner edges of the BGA warp in a concave manner causing loss of contact between the BGA ball and substrate. This was also discovered in previous studies.  Figure 1-5 illustrates where a solder ball had dislodged itself sometime during the reflow processes.

In some cases, as the X-ray in Figure 1-5 illustrates, the solder ball displaces itself as the liquidus stage is reached and coalesces in other areas. This occurs prior to the underfilling stage, even if the underfill is subjected to the reflow process. Once the underfill is cured, especially one formulated with high filler content, the cross linked polymer would prevent the solder from wicking into adjacent areas, since it would be restrained by the underfill.

Wetting
In the majority of the cases where the package sites have continuity, good wetting is evident even among the assemblies utilizing a mixed solder system and a two step reflow process. Figure 1-6 shows an example of good wetting to the substrate and BGA package. Inspection of POP assemblies for opens and shorts utilizing X-Ray techniques presents a bit of a challenge, since the staggered patterns of the top and bottom BGAs can potentially mask defects, or give a false negative reading due to its unusual appearance.

It is sometimes necessary to view the oblique angles for verification of shorts, opens or acceptable ball collapse (Figure 1-7).

Summary
Processing POP assemblies requires carefully controlled measures at each step of the process. Monitoring PCB warpage and BGA ball height will help mitigate the effects of the warpage caused by the CTE mismatch of package and substrate. Applying a controlled amount of tacky flux is also critical to ensuring adequate wettablility to the pads, without creating a bridging effect. The use of underfill and its effect on reliability for POPs is not fully understood yet, but the application of a low CTE underfill is recommended for thermal cycling conditions.