Via-in-Pad Plated over Design Considerations to Mitigate Solder Separation Failure
As signal speeds and performance requirements continue to rise, the use of advanced PCB technologies is becoming increasingly important. As a result, the via-in-pad plated over (VIPPO) structure has been adopted in many BGA footprint designs within the PCB. These VIPPO structures are preferred over the more traditional dog-bone pad structure in order to shrink signal path lengths, reducing two parasitic effects, capacitance and inductance, for improved highspeed performance.
Figure 1 illustrates how the VIPPO structure can influence those parasitic effects. The signal traces, which connect the BGA pads with the vias, act as inductors. Additionally, as high-speed designs typically have ground planes immediately below the outer layer, there is also a capacitive effect that is generated. With the VIPPO structure, the outer trace layer is eliminated, thereby cancelling both parasitic effects.
Figure 1: Dog-bone vs. VIPPO pad structure.
Figure 2 exhibits the VIPPO structure as compared with the VIPPO + backdrill (BD) structure. The use of backdrill with the VIPPO structure can eliminate the reflections within the unused portion of the via, which acts as a stub. The portion of the via indicated by the purple arrow is not in series with the signal path, but instead acts as a stub. Therefore, a portion of the signal is reflected back, creating an interference, which will degrade the high-speed signal performance. Hence, the purpose of the back-drill is to remove this “unused” portion of the via in order to eliminate the reflections for a cleaner signal.
Figure 2: VIPPO vs. VIPPO + backdrill structure.
With increased complexity of PCB designs for high-end networking products, the boards thicknesses are typically >120 mils and signal speeds are reaching 25 GHz and beyond. For these types of designs, backdrilling of the VIPPO structures becomes imperative.
It is also a common practice to mix VIPPO and non-VIPPO pad structures within a single BGA footprint, as indicated in Figure 3. The green lines indicate a high-speed signal trace (e.g., for differential pairs) on the outer layer. It is preferable from a signal integrity perspective, to route these signal lines on the outer layers of the PCB to take advantage of microstrip routing which has faster propagation speeds than stripline routing. Hence, these BGA pads do not require the use of VIPPO. These non-VIPPO pads are highlighted in red. Without any VIPPO structure, a zero stub length can be achieved, which is an extremely attractive option for the signal integrity engineer. Moreover, additional routing space is gained underneath the non-VIPPO pad. Unfortunately, these types of mixed footprint designs have a propensity for manufacturing defects during SMT assembly of BGA packages and can potentially expose the PCBA to field reliability risks if these defects escape manufacturing tests.
Figure 3: Mixed VIPPO/Non-VIPPO BGA Footprint.
Failure Mode and History
As a consequence of these advanced PCB technologies and complex board designs, a unique BGA solder joint failure mode has emerged during specific assembly conditions. This failure mode occurs when the bulk solder separates from the IMC during or just prior to reflow. This failure mode is of particular concern because the discontinuity is so small relative to the size of the solder joint itself that it cannot be detected via X-ray inspection methodologies. Furthermore, in many cases it is only a partial separation of the BGA solder joint and hence, it may not even be detected via ICT or functional test techniques. Without a robust methodology to screen for these defects, this presents an extremely high risk for potential escapes to the field.
Typically, this failure mode has been found on BGA packages with a 1 mm pitch or less BGA array and having a PCB footprint that includes a mixed VIPPO/non-VIPPO pad design. The solder separation occurs when the component is subjected to a secondary reflow, either during top-side SMT for bottom-side components or during rework of an adjacent, or mirrored, BGA component. Since the open occurs between the bulk solder and the IMC, it does not have the typical brittle solder joint fracture signature, which has a flat fracture interface through the IMC as shown in Figure 4. Instead, this failure mode exhibits more of a hot solder tear or separation type of failure mode, as the solder separates from the IMC leaving it intact.
Figure 4: Example of brittle fracture.
Figures 5 and 6 illustrate examples of both partial and complete solder separations. For these failures, the solder separation only occurs on the solder joints that use a VIPPO BGA pad and is typically adjacent to a solder joint(s) with a non-VIPPO BGA pad. In some cases, this type of failure mode has also been identified on a component having a full VIPPO BGA pad pattern on the PCB when there is also a pattern of VIPPO with deep backdrill (BD) within the footprint. Hence, the deep-backdrill VIPPO structures seem to mimic the behavior of the non-VIPPO pads so that it becomes comparable to a mixed VIPPO/non-VIPPO BGA pad footprint and again, induces solder separation in the solder joints on a VIPPO pad when subjected to a secondary reflow.
Since the separated solder ball shape is rounded near the open or partially open interface, this indicates that the solder joint underwent reflow subsequent to the separation. Furthermore, since the separation is between the bulk solder and the IMC and does not reflect a brittle fracture, it is suspected that the separation occurred after the solder has softened and is nearly molten. Figure 6 illustrates a brittle solder joint failure in which the fracture occurs within the IMC itself and exhibits more of a flat surface indicative of crack propagation.
Figure 5: Partial solder separation.
There seem to be two primary effects that are occurring which contribute to this failure mechanism. First, there is a CTE mismatch between the VIPPO structure and the non-VIPPO, or deep-backdrill VIPPO, structure, that results in a greater expansion of the PCB beneath the non-VIPPO BGA pad, or the deep-backdrill VIPPO pad, as compared with the VIPPO BGA pad. Secondly, the higher thermal conductivity of the VIPPO structure as compared with the non-VIPPO, or deep-backdrill VIPPO structure, allows the VIPPO solder joints to reach liquidus before the adjacent solder joints having a non-VIPPO, or deep-backdrill VIPPO pad. Therefore, during a secondary reflow process, when the adjacent non-VIPPO solder joints are still solid, tensile stresses are induced in the VIPPO solder joints as the adjacent non-VIPPO solder joints are pushed up due to the greater out-of-plane PCB expansion beneath those pads.
Subsequently, when the VIPPO solder joints become molten, these high stresses are relieved as the bulk solder “tears” or separates from the IMC. This solder separation can occur at either the package interface or the PCB interface of the solder joint, depending on whichever is the weaker interface. Since the PCB pad design is generally a non-soldermask-defined pad (NSMD) and the BGA package typically uses soldermask-defined pads (SMD), the separation will more likely occur at the package interface.
Figure 6: Complete solder separation.
Alternatively, a 100% VIPPO BGA footprint without deep backdrill does not introduce the additional stresses that are exhibited with the CTE mismatch between adjacent VIPPO and non-VIPPO pad designs. Additionally, a 100% VIPPO BGA footprint without deep backdrill does not create the high thermal gradients between adjacent solder joints that the mixed VIPPO/non-VIPPO BGA footprints achieve. Therefore, this type of failure mode has not been identified with 100% VIPPO BGA footprints with no deep backdrill.
In order to better understand the influence of various PCB and packaging design parameters on this failure mode, three different test vehicles have been designed to assess the following:
1. Influence of drill hole size (DHS) for the VIPPO structures: 9.8 mils vs. 7.9 mils DHS
2. Influence of BGA package body size and BGA pitch
3. Influence of varying backdrill (BD) depths and BGA package body size
Each test vehicle is assembled through a primary and secondary SMT attach process, followed by inspection and physical analysis to validate the solder joint integrity after each reflow. The printed circuit assembly equipment, process parameters, tooling (e.g., stencil design and technology), assembly materials (e.g., solder paste) and inspection equipment and methodologies utilized for these builds are consistent with Cisco’s standard production processes in order to minimize the number of variables introduced in this study.
To read the full version of this article, which appeared in the November 2017 issue of SMT Magazine, click here.