Rework and Reball Challenges for Wafer Level Packages
There is a growing consumer demand for smart wearable products, including watches, fitness bands, eyewear, and headphones. Barriers to the market still exist, as many consumers are concerned with style, battery life, functionality, and cost. The mature smartphone market faces similar challenges, with consumers desiring thinner phones with longer battery life and increased sensor functionality. High-density and stacked PCB designs, system in package (SiP) assemblies, and small form factor packages have all emerged as solutions for sleeker styles and improved functionality. In particular, WLPs are gaining popularity for their low cost, small footprint, and thin profile. The newest smartphone models contain an average of 5-7 WLPs, with many WLPs used as RF transceivers, power management units, audio amplifiers, and BlueTooth and GPS modules.
Figure 1: Cross-section schematics of (a) a traditional fan-in wafer level package, and (b) a newer fan-out wafer level package.
Unlike traditional packages, WLPs are packaged and bumped first, then diced. Passivation and dielectric layers are added to the die frontside, followed by metallic redistribution layers. A second dielectric layer is deposited, then the underbump metallization and solder balls are attached. Lastly, the packages are singulated from the wafer. Figure 1 illustrates the two different categories of WLPs: the traditional a) fan-in WLP, and the newer b) fan-out WLP. The dielectric is exposed on the edges and frontside of fan-in WLPs, while the silicon backside is often covered with a protection tape. As the name suggests, fan-out WLPs “fan out” interconnects from the smaller silicon die to the larger package dimensions. The fan-out design allows for ball pitch customization, higher I/O density, and easy integration with SiPs and other multidie packages. The board-level reliability is also improved by protecting the silicon die with an epoxy mold compound.
Though WLPs offer considerable advantages, they also pose challenges for failure analysis—particularly when reball and component-level testing are required. Figure 2 illustrates the typical FA process flow for a failing system, such as a mobile phone, tablet, or wearable device.
Figure 2: Diagram showing the typical process flow for system-level failure analysis
Fault isolation is performed first to identify the failing component, then nondestructive FA is used to inspect for failures at both the board- and package-level. If the failure is subtle or cannot be found non-destructively, the package must be reworked and reballed before proceeding with socketed component-level testing and further FA. Due to their small form factor and fragility, WLPs can prove particularly difficult for the rework and reball process steps. The present work provides an overview of the FA process flow, emphasizing component rework and reball methodologies. The challenges of a specific WLP case study are discussed, and rework and reball improvements are implemented to minimize thermally- or mechanically-induced artifacts.
Fault isolation is an important first step in the system-level FA process. By isolating the failure to a specific package or interconnect, an optimal FA approach can be assessed and throughput time can be greatly reduced. Several fault isolation techniques can be used to measure for opens and shorts, including hand probing, time domain reflectometry and the newer electro optic terahertz pulse reflectometry. High-resistance shorts can also be detected by powering the board and measuring the localized temperature increase using infrared thermal imaging techniques.
After the failing component has been isolated, nondestructive FA is performed to inspect for gross board- or package-level failures. OM can be used to inspect for external board or package defects, such as foreign materials or superficial die and overmold cracks. Since WLPs have exposed dielectric layers (and since fan-in WLPs have exposed bulk Si), it is especially important to perform a first pass inspection for external chips or cracks. It is recommended that optical inspection be performed again after rework and reball, to confirm that no artifacts were introduced.
CSAM, a popular non-destructive technique, can be utilized to detect internal defects or to evaluate the extent of external damage. CSAM uses an ultrasound transducer to raster-scan the package backside. At material interfaces, an acoustic pulse is reflected to the transducer and recorded as signal amplitude. Air-solid interfaces occur at the locations of cracks, voids, or delamination, and return high-intensity reflections. CSAM is thus a valuable metrology for identifying gross internal package defects that cannot be detected with simple optical inspection. CSAM can also be used to screen for rework and reball artifacts, particularly thermally- induced delamination.
2D X-ray is another common technique that provides an effective “quick pass” inspection for board-level solder defects, such as voids and bridging. By optimizing the sample tilt and rotation angles, subtler non-wet open and noncontact open defects can also be detected. However, 2D X-ray is not capable of detecting submicron defects, such as board-level solder or via cracks.
Recently, more advanced imaging metrologies have emerged as powerful non-destructive FA techniques. In particular, 3D X-ray computed tomography has proven effective at detecting both board-level and package-level sub-micron defects. Multiple 2D X-ray images are collected as the sample is rotated at fixed angle increments. The 2D X-ray images are then superimposed to generate a three-dimensional volume. The superimposed image can be manipulated to display virtual “slices” of the sample, allowing for inspection of the solder joints, via barrels, and traces.
If board-level or gross package-level failures are not detected using non-destructive techniques, the package must be sent for socketed component-level testing and further FA. Standard test sockets use a spring-loaded floating base that is guided by the solder balls instead of the package edge. Reball is thus required to align the package and enable good electrical contact with each pin. It is very important to preserve the defect signature prior to testing; accordingly, precautions must be taken to improve the reball yield and reduce thermal and mechanical artifacts.
Rework and reball process yield is influenced by several different factors, and can be classified into four general categories, including: (1) personnel; (2) methods; (3) materials; and (4) machine/ tooling. Figure 3 summarizes the various categories and sub-categories for a typical PCB assembly.
Figure 3: Diagram showing the factors that dictate the yield of rework and reball processes.
The personnel category represents the “human factor,” and includes handling, training, and quality control. The personnel factor is especially important for WLPs and other small form factor devices, as the components are much more fragile than standard flip-chip packages. Proper handling must be used to minimize mechanical artifacts and prevent damage to the bulk Si or dielectric layers. The methods category encompasses the process steps for rework and reball, including sample preparation, package removal from the board, solder removal from the package-side pads, and reball of the package. Prior to demount, the sample must be prepared by removing any heat spreaders, thermal grease, corner glue, or underfill. The package is then demounted from the PCB using either mechanical or thermal methods. Next, solder is removed from the package-side pads using solder wicking with a braided wire and solder tip, or using a no-contact vacuum scavenging technique. Lastly, the package is reballed using either a stencil, preform, or laser jetting method.
Materials—such as package type, PCB design, flux, underfill, and corner glue—also influence the rework and reball process yield. With higher board densities and package miniaturization, it becomes increasingly difficult to selectively heat and remove parts from a small footprint. Since WLPs do not possess a package substrate or solder mask, there is also an increased risk to damage the dielectric or even the metal redistribution layers.
Lastly, rework yield is impacted by the machines and tooling used to handle and process the package. Rework machines can vary greatly in cost and complexity—ranging from a hot air pencil and tweezers, to a fully automated rework station. Semi-automated and fully automated rework tools are expensive, but can greatly minimize the risk for thermal and mechanical artifacts. The latter is especially important for WLPs, as it is difficult to handle and secure the parts.
To read the full version of this article, which appeared in the September 2017 issue of SMT Magazine, click here.