The future of electronic package is 3D. From stacked die to Intel’s recent introduction of the 3D transistor1, the continued progression in Moore’s Law is being accomplished by going ‘up’. For the majority of design engineers, nowhere is this trend more concrete then the increasing utilization of components offered with a package-on-package (PoP) architecture.
Package on package is a configuration where two packaged integrated circuits are placed directly on top of each other (aka, stacked packages). The interconnects, or solder joints, are between the top package and the bottom package and the bottom and the PCB (note, some PoP configurations have more than two packages in the stack2). The top package is traditionally thicker, containing multiple or stacked die, while the bottom package is thinner, with either smaller or thinner die. While various patents, and some commercial product, have focused on PoP with leaded devices, the overwhelming configuration is with ball grid array (BGA) technology. Implementation of the PoP concept took off in the 2003-2005 timeframe, with the first JEDEC package outlines defined in 20063.
As with all new technology4 introduced into the electronics marketplace, the prevalence of PoP has been greatly exaggerated. PoP recent ascendancy has been primarily been through a few high volume manufacturers in the mobile and memory space. However, as PoP becomes a more common solution, it will become important for companies in low volume, high rel applications to understand the technology’s limitations in regards to manufacturability and reliability.
The benefits of PoP are well known. They include
To experience the benefits of PoP, the risks of PoP, especially in regards to manufacturability and reliability, must be effectively managed. These two requirements cannot always be completely separated and therefore must be managed concurrently. Most critically, the designer must realize when and how to modify standard BGA packaging design or process guidelines.
One of the greatest challenges in assembling PoP is managing out-of-plane warpage5. For the purposes of this paper, it will be assumed that the OEM has no control over the design of the PoP construction. Therefore, the OEM must somehow deal with the warpage that is intrinsic with the PoP delivered by the part manufacturer. The bottom PoP packages are especially susceptible due to their construction, thin body, and low-profile solder balls. The direction of warpage, concave or convex, is also not necessarily intuitive and is based on a complex interaction between the die, molding compound, and substrate (and, to a lesser extent, the underfill/die attach).
Therefore, a critical first step in any DfM / DfR activity is to characterize6 the warpage behavior of the PoP part over the expected reflow profile. Within the bounds of a nominal reflow profile (ramp rates, hold times, peak temperature), the warpage is not expected to vary dramatically7. The primary control the OEM can have over PoP warpage is to select either a one-pass or twopass8 assembly.
One pass assembly involves mounting and reflowing the top and bottom packages at the same time. The bottom package is first placed on the board. The top package is then dipped in flux or solder paste and then placed on the bottom package. Both packages are then run through the reflow oven in a single pass. For a two pass assembly, the top and bottom packages are assembled in a separate process (either at the outsourced semiconductor assembly and test OSAT) manufacturer or at the contract manufacturer (CM)).
While there are logistical and cost drivers that influence one or two-pass assembly, the influence of the stackup on warpage behavior should be taken into consideration. This is especially true if the measured warpage during one-pass, the preferred assembly method, is close to or exceeds industry specifications (see Table 1). Industry best practice is to typically aim for 0.05 mm of warpage at the reflow temperatures (or, at least, less than the coplanarity requirements). The actual warpage of the PoP, in terms of separation and shape (convex or concave) can potentially have a strong influence on bond pad design.
The warpage of the PoP will play a key role in determining if the board bond pads will be solder mask defined (SMD) or non-solder mask defined (NSMD). This is a surprisingly complex decision and there are a number of advantages and disadvantages to both approaches in regards to manufacturability and reliability.
NSMD is the traditional board bond pad design for BGA packages and there are several advantages to this approach. A NSMD solder joint will typically last longer than a SMD solder joint when subjected to temperature cycling and vibration because of the additional attachment around the edge of the bond pad. A NSMD pad is also smaller, allowing more space for routing of traces between pads, and provides more uniformity because the copper etching process is more controlled than solder mask. However, finer pitch offerings (down to 0.3mm) start to make it difficult to maintain the soldermask webbing.
SMD pads allow for a thicker web, but less room for routing. SMD bond pads have a larger surface area attachment to the laminate, which can prevent pad cratering that can occur due to excessive flexure during manufacturing or operation (e.g., drop). SMD solder joints also have a higher standoff. This provides greater compliance, making for a more robust solder joint during shock or drop events13, and allows easier access for cleaning. It is the ability of SMD to provide a higher standoff that can be crucial in mitigating the out-of-plane warpage experienced in PoP packages.
Because SMD has less of a collapse than NSMD, some companies have implemented SMD bond pads on PoP packages as a way to mitigate warpage. When faced with more challenging warpage behavior and tight production deadlines, a non-standard solution has been a hybrid of NSMD and SMD to temporarily maximize assembly yield. An example of resolving a difficult concave warpage in a PoP is to populate the peripheral bond pads with a SMD design with a minimum solder mask opening and to slowly open the solder mask to a standard NSMD bond pad in the center of the array.
Other bond pad design attributes tend to have a more moderate effect on manufacturability. Board bond pad diameters should be equivalent to package bond pad dimensions (typically about 70% of the ball diameter; see Table 2). Smaller bond pad diameters can reduce the degree of collapse, but will also introduce asymmetry which can greatly reduce time to failure in temperature cycling and vibration environments.
Few, if any studies, are available on the effect of bond pad geometries on PoP manufacturability. Some BGA designs currently use a non-circle geometry, such as square or octagon, to improve manufacturability and this same modification may provide some value in a PoP assembly process.
Of course, solder mask and bond pad design is just one aspect of PoP design for manufacturability. Stencil design and solder materials also play a critical role.
Many component manufacturers and other sources recommend that for PoP assembly the stencil aperture should be slightly larger than the bond pad to induce an overprint. This provides a slightly better paste release and, more importantly, increases the volume of solder paste to mitigate any coplanarity / warpage issues. This approach, which should only be used for NSMD pads, can be challenging as it can lead to solder splash, solder balling, and potential bridging, especially for fine pitch designs.
For a lead-free process, the typical aperture / bond pad ratio is 1:1. There is some debate within the electronics community about the geometry of the aperture. Some prefer a square aperture for its release characteristics, especially for smaller openings. Others will claim round apertures provide a more uniform deposit of paste. Some design of experiments may be required, especially at 0.4mm and 0.3mm pitch.
Finally, the choice of assembly materials is just as an important DfM activity as board-based geometries. The paste used for the interconnection between the bottom package and the printed board should be optimized for slump and wetting characteristics, especially at elevated temperatures. Anti-head in pillow formulations are recommended given PoP susceptibility to this phenomenon. For the connection between top and bottom package, there is a choice of paste or tacky flux. Tacky flux is typically preferred in high volume applications. When applied, the coverage of the flux should be to at least half-way up the solder ball. The application of solder paste is more challenging (screen printing is typically not an option), but provides more margin in terms of warpage / coplanarity issues.
This white paper may include results obtained through analysis performed by DfR Solutions’ Sherlock software. This comprehensive tool is capable of identifying design flaws and predicting product performance. For more information, please contact DfRSales@dfrsolutions.com.
DfR represents that a reasonable effort has been made to ensure the accuracy and reliability of the information within this white paper. However, DfR Solutions makes no warranty, both express and implied, concerning the content of this report, including, but not limited to the existence of any latent or patent defects, merchantability, and/or fitness for a particular use. DfR will not be liable for loss of use, revenue, profit, or any special, incidental, or consequential damages arising out of, connected with, or resulting from, the information presented within this white paper.
To maximize Design for Manufacturability (DfM) and Design for Reliability (DfR) in order to mitigate risk, Sherlock evaluates key components, including: Solder joint reliability to assure a product will function under given conditions, for a specified time, without exceeding defined failure levels, plated through hole fatigue by using computerized modeling and temperature maps instead of human interfaces for accurate finite test results.
Failure is a possibility with any component on any PCB. In many cases wearout is the culprit, leaving engineers to deal with the aftermath of dissecting what went wrong and possibly re-engineering the component to avoid recurrence.Failure is a possibility with any component on any PCB. In many cases wearout is the culprit, leaving engineers to deal with the aftermath of dissecting what went wrong and possibly re-engineering the component to avoid recurrence.